Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Disenrollment from general practitioners among chronic patients: a register-based longitudinal study of Norwegian claims data

Disenrollment from general practitioners among chronic patients: a register-based longitudinal... Background: Norwegian general practitioners (GPs) consult on a variety of conditions with a mix of patient types. Patients with chronic diseases benefit from appropriate continuity of care and generally visit their GPs more often than the average patient. Our aim was to study disenrollment patterns among patients with chronic diseases in Norway, because such patterns could indicate otherwise unobserved GP quality. For instance, higher quality GPs could have both a greater share of patients with chronic diseases and lower disenrollment rates. Methods: Data on 384,947 chronic patients and 3,974 GPs for the years 2009–2011 were obtained from national registers, including patient and GP characteristics, disenrollment data, and patient list composition. The birth cohorts from 1940 and 1970 (146,906 patients) were included for comparison. Patient and GP characteristics, comorbidity, and patient list composition were analyzed using descriptive statistics. Patients’ voluntary disenrollment was analyzed using logistic regression models. Results: The GPs’ proportion of patients with a given chronic disease varied more than expected when the allocation was purely random. The proportions of patients with different chronic diseases were positively correlated, partly due to comorbidity. Patients tended to have lower disenrollment rates from GPs who had higher shares of patients with the same chronic disease. Disenrollment rates were generally lower from GPs with higher shares of patients with arthritis or depression, and higher from GPs who had higher shares of patients with diabetes type 1 and schizophrenia. This was the same in the comparison group. Conclusion: Patients with a chronic disease appeared to prefer GPs who have higher shares of patients with the same disease. High shares of patients with some diseases were also negatively associated with disenrollment for all patient groups, while other diseases were positively associated. These findings may reflect the GPs’ general quality, but could alternatively result from the GPs’ specialization in particular diseases. The supportive findings for the comparison group make it more plausible that high shares of chronic patients could indicate GP quality. Keywords: Chronic patients, Switching, Primary health care, Schizophrenia, Epilepsy, Diabetes type 1, Diabetes type 2, Asthma, Arthritis, Depression * Correspondence: anastasiya.mokienko@gmail.com; anastasiya.mokienko@gmail.com Department of Health Management and Health Economics, University of Oslo, P.O. Box 1089, Blindern, Oslo 0318, Norway © The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 2 of 10 Background If patients switch between GPs until their demands are The quality of care for people with chronic diseases met, we would expect these patients to be dispropor- often relies on appropriate primary care. Some such tionally distributed across GPs. Similar trends could be patients may need continuous, long-term follow-up and expected if the GPs intentionally specialize, formally or motivation in order to maintain a favorable lifestyle. not, in a given patient group. However, neither of these Others, who experience a condition associated with mechanisms has obvious implications for the provider social stigma, may need time to develop trust in their choices made by other groups of patients. For example, care providers. Early detection of the chronic disease a GP who is popular among patients with diabetes type and its subsequent routine monitoring is also very 2 (DT2) may also be popular among patients with important to save patients from acute hospitalization depression, whereas patients without chronic diseases and complications from the disease [1]. Comorbidity is a may be indifferent to this GP’s motivational skills. Older good reason for primary care providers to be better able patients and patients with chronic diseases have generally to manage chronic diseases [2, 3]. higher care continuity, whereas patients with lower care Previous studies have found that long-term physician- continuity are those living in rural areas, employed, with patient relationships are beneficial for patients [4, 5] and higher education, or with poorer mental health [26]. that patients disenroll from their general practitioner Our aim is to investigate patterns of chronic patient (GP) when they are not satisfied with their GP-patient disenrollment. This type of study is required because relationship [6–10]. Patients may also disenroll from their there are no published indicators of GP quality, and GP if they perceive insufficient quality of care. Accessibility therefore these indicators need to be identified through factors, such as adequate time for consultations [11] and patient actions (such as disenrollment). Moreover, spe- availability of appointments [12] are predictors of good cialized patient choice patterns might suggest an extra quality. Booking intervals for consultations and duration argument for using more fee-for-service reimbursement of the consultations themselves are correlated with good or risk-adjusted capitation for GPs in order to compen- management of chronic diseases; the effect was greater for sate for varying expected workloads depending on their patients with asthma than for those with diabetes or patient list composition. Primary care in Norway is pub- angina, possibly because primary care providers deal licly funded with a capitation and fee-for-service system, more with asthma than diabetes or angina [13]. and patients have to consult their GPs in order to see a When it comes to accessibility, earlier research shows specialist. Each individual GP has a patient list and can that longer patient lists are associated with negative decide the maximum number of patients that can be evaluations of accessibility and that the GP's age has a enrolled on their list. Patients can switch between available negative association with the evaluation of all aspects, GPs up to three times a year, according to their own except accessibility [14]. Longer patient lists are also asso- preference. ciated with better illness detection [15], which may sug- gest that practices detecting a higher number of chronic Methods conditions have greater demand from patients due to their Data sources and study populations systematic chronic disease management [15–18]. This is a retrospective study using data from two national A strong connection between patient choice and registers in Norway, administrated by the Norwegian higher quality of practice, as measured by studying the Directorate of Health, from 2009–2011. Our GP data publicly available data on practice performance, has been were obtained from the national register of regular reported [19]. A review study found that patients were GPs, which covers the entire GP population, and weakly influenced by publicly available information merged with patient data using the GPs’ IDs. Our patient about provider quality [20]. On the provider side, only data were based on claims data obtained from the KUHR hospitals seemed to improve quality as a response to registry (Kontroll og Utbetaling av HelseRefusjon), which quality indicators being made publicly available [21]. For covers the entire Norwegian patient population. This GPs, patient shortage has been found to correlate with registry records claims data continuously but for our ana- patient dissatisfaction, the GP’s communication skills, lysis, the sample period 2009–2011 was divided into six and other GP characteristics [22–24]. semiannual intervals. The individual level data included Interaction between chronically ill patients and their patient characteristics, their consumption of primary care, GPs has not been given specific attention in previous and the GP with which they were enrolled. literature, but a previous study of obese patients may Two samples of patients were selected among patients contain clues for generalizable results: reportedly, obese who visited a GP at least once from 2009–2011. Most of patients avoided physicians they perceived as sources of our analysis is based on sample 1, which consisted of pa- stigma and searched for providers who were “obese tients registered with one or more of the following seven friendly” [25]. diagnoses at least once during the period 2006–2011: Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 3 of 10 DT1, DT2, asthma, arthritis, schizophrenia, depression, Measures and epilepsy. These patient groups were chosen because Our main outcome variable, ‘SwitchOut’, measured they are known to vary substantially both in the number whether a patient disenrolled from a GP from one semi- of patients in the population, and in the utilization of annual period to the subsequent period. Definitions of primary care services. For instance, patients with DT2 independent variables are summarized in Table 1. Infor- constitute almost 5% of the population and receive most mation about the GPs’ age, sex, specialization, and list of their health care from their GP, while patients with length, and patients’ sex, birth year, and number of visits schizophrenia are fewer and receive more specialist care was obtained directly from the data registries. The vari- in a hospital setting. able ‘Pat_comorb’ was given the value 0 for patients in Our analysis also included a comparison group, sample 2. sample 2, while for each patient in sample 1 we counted This group consisted initially of the entire birth year the number of registered diseases (1–7) and subtracted 1 cohorts from 1940 and 1970, but we excluded patients from this number. This yielded a variable with a range be- already included in sample 1. Obviously this selection tween 0 and 6. The variables ‘Diab2_share’ and ‘Epil_share’ yielded an age distribution different from that in sample measure a GP’s share of patients with the respective 1, but the selection of one elderly and one younger chronic disease, but with a slight adjustment: if shares birth year cohort should provide a good basis for were calculated straightforwardly, they could potentially comparison. be influenced by the health status of a single patient, Initially, the two samples combined contained 988,483 because some chronic diseases are relatively rare and patients (Fig. 1). We excluded 34,189 cases where the some GPs had fewer patients (shorter lists). To illus- disenrollment was likely to be due to causes not relevant trate, consider a GP who has 100 patients, of which one for our purpose; that is, when patients moved to another has epilepsy. If we take the perspective of the GP, the municipality, or when a GP moved, retired, or died. For share of patients with epilepsy is slightly above average the logistic regressions, we excluded patients living in (Table 1). However, this measure is of little relevance if municipalities with less than 50,000 inhabitants in order we take the perspective of the patient with epilepsy: the to focus on patients who could choose from several GPs. GP has no other patients with epilepsy. To avoid inter- This left us with 316,636 patients in sample 1 and 32,311 pretational ambiguity, we chose to take the patients’ patients in sample 2 (348,947 in total). Finally, we perspective. For each patient-GP pair, we excluded the excluded patients with irregular medical records, mainly patient from the calculation of the GP’s share. Thus, the missing birth year or sex, yielding 313,659 patients in sam- share variables mostly showed the variation between GPs ple 1 and 30,212 patients in Sample 2 (343,871 in total). but also some variation within a GP practice. Fig. 1 Flow diagram of sample selection Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 4 of 10 Table 1 Variable definitions and descriptive statistics on the patient level Variable Definition Sample 1 (N = 313,659) Sample 2 (N = 30,212) Median Mean St.dev Median Mean St.dev DT1_share The share of a GP’s patients with diabetes type 1 0.005 0.007 0.006 0.005 0.006 0.005 DT2_share The share of a GP’s patients with diabetes type 2 0.042 0.046 0.027 0.036 0.040 0.021 Arth_share The share of a GP’s patients with arthritis 0.014 0.016 0.010 0.013 0.015 0.009 Asthm_share The share of a GP’s patients with asthma 0.020 0.023 0.015 0.018 0.020 0.013 Depr_share The share of a GP’s patients with depression 0.107 0.112 0.042 0.094 0.100 0.038 Schi_share The share of a GP’s patients with schizophrenia 0.004 0.005 0.003 0.004 0.005 0.003 Epil_share The share of a GP’s patients with epilepsy 0.008 0.009 0.004 0.008 0.008 0.004 ListLength The number of patients on a GP’s list 1423 1444.0 367.8 1439 1453.4 367.8 Ln_ListLength The natural logarithm of Listlength 7.261 7.240 0.277 7.272 7.248 0.270 GP_Age The GP’s age 52 50.358 9.120 51 49.744 8.989 GP_Sex =1 if the GP is male, =0 otherwise 1 0.706 0.455 1 0.673 0.469 GP_age · GP_Sex The product of GP_Age and GP_Sex 48 36.473 24.718 45 34.266 25.016 GP_Specialist =1 if the GP has a specialist degree 1 0.707 0.455 1 0.702 0.457 in general medicine; =0 otherwise Pat_Sex =1 if the patient is male; =0 otherwise 0 0.426 0.494 0 0.494 0.500 Pat_BirthYear The patient’s year of birth 1959 1958.6 19.1 1970 1961.5 13.5 Pat_Comorb Sample 1: No. of chronic diseases minus one. 0 0.148 0.405 - Sample 2: Not defined Pat_Visits The patient’s number of visits to primary care 3 4.662 5.268 1 2.227 3.369 th Pat_Visits_win Winsorized Pat_Visits at 99 percentile (max = 23) 3 4.570 4.626 1 2.205 3.107 Pat_Visits_dum =1 if Pat_Visit >23, =0 otherwise 0 0.10 0.98 0 0.002 0.047 Municipalities over 50 000. First half of 2009 In order to avoid highly influential outliers, we trans- sub-samples partly overlapped due to comorbidity. For formed two variables. The distribution of GPs’ list length each sub-sample, the shares of patients with 1 of the was skewed so we transformed the variable using the other six diseases were calculated. natural logarithm. The distribution of patients’ number We then used logistic regressions to model patients’ of visits to primary care was also skewed, and for this disenrollment from their GP. The modeling was per- th variable, we winsorized the distribution at the 99 per- formed for each patient category separately: on the sub- centile (23 visits per period) and included a dummy vari- samples from sample 1, as defined above, and sample 2. able for observations that exceeded this limit. Because the dependent variable (SwitchOut) was based on observations from two consecutive periods, we had Statistical analyses up to five effective observations for each patient. For the We inspected the data numerically and graphically at independent variables, we used observations from the both the patient and GP levels. This included graphs first five periods. The set of independent variables intended to reveal whether the distribution of chronic included those from Table 1, and an interaction term patients seemed disproportionate across GPs. On the GP between GPs’ age and sex. We incorporated the longitu- level, the mean proportion of patients with DT2 was dinal data structure by including patient-specific effects 4.5% in the first half of 2009. If patients were allocated (intercepts) in the models. Patient-specific effects can by pure chance, a randomly selected GP’s share of account for unobserved factors, such as ethnicity or patients with DT2 would have the expected value of educational background, as long as these factors remain about 4.5%, and be approximately normally distributed constant throughout the sample period. The models were for a sufficiently long patient list (>60 patients). For data estimated using xtlogit in Stata 13, under the standard at the GP level, we calculated Spearman’s correlation co- assumptions that the patient-specific effects were nor- efficients for the various GP-related variables, including mally distributed and did not correlate with the inde- the shares of patients with different diagnoses, the GP’s pendent variables. Fixed effect models, which allow the age and sex. We defined sub-samples of patients from patient-specific effects to be non-normally distributed or sample 1 based on the seven chronic diseases. These correlated with the independent variables, were also Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 5 of 10 considered. However, in fixed effect models the time- Table 2 Share of patients who voluntarily disenrolled from their st nd 1 GPs, between the 1 and 2 halves of 2009. invariant patient variables for sex and birth year would, by construction, be excluded from the estimations. Sample Sub set N % Sample 1 Full sample 313,659 4.52 Results DT1 11,292 4.99 Descriptive statistics DT2 74,473 3.75 According to Table 1 and Fig. 2, the proportion of patients Schizo 8,316 6.29 with DT2 varied substantially among GPs. If these patients Depr 186,415 5.00 had been allocated purely by chance, about 95% of the Arthr 27,157 4.00 proportions would lie between the red curves in Fig. 2, but this was not the case. In fact, only 46.5% of the pro- Asthm 37,110 4.16 portions were positioned within the red curves. For the Epil 15,403 4.86 other diagnosis groups, the corresponding patient shares Sample 2 Full sample 30,212 3.76 also seemed disproportionally distributed. Municipalities over 50 000 Overall, 4.5% of chronic patients disenrolled from their GP from one period to the next, but the share varied from 3.7% among patients with DT2 to 6.2% among was the least frequent (N = 21,368). In the sub-sample of patients with schizophrenia (Table 2). Among patients in patients with depression (third column from the left), sample 2, the share that disenrolled was 3.7%. 1.3% also suffered from schizophrenia. Among patients Descriptive statistics for the independent variables with schizophrenia (rightmost column), 28.7% also suf- used in the logistic regressions are reported in Table 1, fered from depression. A substantial number of patients separately for samples 1 and 2. Due to the sample selec- were recorded with both DT1 and DT2, likely due to tion procedure, the average GP characteristics differ some- registration errors or diagnostic uncertainty. what from those obtained for the full GP population, We calculated Spearman’s rank correlation coefficients where 66% were men, the average age was 48 years, and for the GP proportion of patients with a given chronic the average patient list length was 1200 (N = 3940). disease and other patient proportions and GP charac- The distribution of the variable ‘ListLength’ appeared teristics, as shown in Table 4. The correlation coeffi- continuous but was somewhat skewed to the right. The cient of ‘Asthm_share’ and ‘DT1_share’ was 0.648, distribution of ‘Pat_visits’ was markedly right-skewed, and indicating that GPs with a high proportion of patients the distribution’s tail was rather scattered: for sample 2, with asthma also tended to have a high proportion of th the 75th, 95 , and 99th percentiles were 6, 14, and 23, patients with DT1. All variables related to the GPs’ pro- respectively, but the maximum value was as high as 219. portions of patients were significantly different from Table 3 presents the sizes of the sub-samples defined for zero. The proportion of patients with chronic diseases the seven chronic diseases. The most frequent of the dis- were all positively correlated, and negatively correlated eases was depression (N = 488,686), while schizophrenia with the proportion of other patients (‘Other_Share’). ‘Other_Share’ was negatively correlated with ‘GP_Age’ and ‘GP_Sex’, indicating that older GPs and male GPs tended to have fewer patients without our seven chronic diseases. Logistic regression analysis Table 5 shows the estimated parameters of the logistic regressions where ‘SwitchOut’ is the dependent variable, the independent variables are those listed in Table 1, and Sigma_u denotes the standard deviation of the patient- specific intercepts. The first seven columns show results based on sample 1 according to patient diagnosis group; the last column is based on sample 2. In logistic regres- sions, the coefficients can be used to compare the differ- ence in log-odds ratios between groups, so that a patient Fig. 2 Scatter plot of GP proportion of DT2 patients and patient list sex coefficient of −0.188 (arthritis patients) represents the length. Legend: Y-axis percent of DT2 patients, X-axis patient-list difference in log-odds ratios between male and female length. GP level, data for the first quarter of year 2009, N = 3,965, patients. The corresponding difference in odds ratios is mean proportion of DT2 patients = 0.045, patient-list lengths of >60 obtained by taking the anti-log, exp(−0.188) = 0.829. Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 6 of 10 Table 3 Percent of patients with a chronic disease (column) that have another chronic disease (row) Arthritis Asthma Depression DT2 DT1 Epilepsy Schizophrenia Arthritis 4.4 2.7 3.9 4.0 2.0 1.3 Asthma 6.1 4.5 6.7 5.8 3.6 6.0 Depression 14.5 17.6 13.5 15.2 15.8 28.7 Diabetes type 2 10.0 12.4 6.4 77.8 5.7 12.0 Diabetes type 1 1.6 1.7 1.1 12.0 1.3 1.9 Epilepsy 1.0 1.3 1.5 1.1 1.6 3.1 Schizophrenia 0.3 1.0 1.3 1.1 1.1 1.4 N 90,095 124,776 488,686 232,383 35,887 46,145 21,368 First half year of 2009. Patient level data. Sample 1 without restrictions (neither on municipality size, data irregularity or moving). N is the number of patients with the chronic disease The statistical inference for this type of model is based disenrollment from GPs with relatively high shares of pa- on large-sample theory and coefficient estimates are tients with arthritis. For ‘Asthm_share’ and ‘Depr_share’, approximately normally distributed. Thus, to simplify all of the significant coefficients were also negative. In the presentation, we do not report p-values as they can contrast, for ‘DT1_share’, ‘Epil_share’ and ‘Schi_share’, be derivedfromthe estimatedstandarderrors. almost all significant effects were positive. Some of the estimated effects of the patient share vari- We can distinguish two main effects. First, the “own ables were relatively robust across patient groups. For share effect,” namely, all patient groups tended to remain ‘Arth_share’, all coefficients were significantly negative, with GPs who had a high share of patients with the same implying that all patient groups tended to have lower diagnosis. Second, the “cross share effect,” where, for Table 4 GP characteristics. Spearman’s correlation coefficients with two-sided p-values. Arth_ Asthm_ Depr_ DT1_ DT2_ Epil_ Schi_ Other_ GP_ GP_ List share share share share share share share share age sex Length Asthm_share 0.488 0.000 Depr_share 0.195 0.264 0.000 0.000 DT1_share 0.519 0.648 0.221 0.000 0.000 0.000 DT2_share 0.232 0.310 0.121 0.332 0.000 0.000 0.000 0.000 Epil_share 0.270 0.298 0.205 0.335 0.177 0.000 0.000 0.000 0.000 0.000 Schi_share 0.045 0.175 0.227 0.135 0.183 0.162 0.005 0.000 0.000 0.000 0.000 0.000 Other_share −0.562 −0.683 −0.762 −0.712 −0.362 −0.406 −0.285 0.000 0.000 0.000 0.000 0.000 0.000 0.000 GP_Age 0.203 0.137 0.064 0.213 −0.047 0.091 −0.028 −0.174 0.000 0.000 0.000 0.000 0.003 0.000 0.077 0.000 GP_Sex 0.181 0.293 0.077 0.318 0.101 0.205 0.135 −0.265 0.249 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ListLength −0.020 −0.069 0.041 −0.032 −0.145 −0.040 −0.033 0.035 0.166 0.172 0.205 0.000 0.010 0.046 0.000 0.011 0.038 0.026 0.000 0.000 GP_Specialist 0.008 0.017 0.030 0.037 −0.133 0.067 −0.003 −0.018 0.365 0.098 0.226 0.618 0.275 0.063 0.020 0.000 0.000 0.860 0.250 0.000 0.000 0.000 GP level data for first quarter of 2009, N = 3974. Correlation coefficients with two-sided p-values less than 1% are in boldface Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 7 of 10 Table 5 Logistic regression for patients’ voluntary disenrollment from GPs, separate for patient groups. Estimated parameters (standard errors) Arthritis Asthma Depression Diabetes type 2 Diabetes type 1 Epilepsy Schizophrenia Others Arth_share −15.032 −10.550 −16.792 −9.506 −16.905 −16.495 −20.113 −15.310 (1.611) (1.597) (0.815) (1.194) (3.116) (2.836) (3.925) (2.185) Asthm_share −4.381 −10.406 −2.117 1.883 −1.624 −0.188 −3.895 0.093 (1.598) (1.309) (0.636) (0.934) (2.494) (2.262) (2.922) (1.799) Depr_share −1.915 −2.343 −5.377 −2.781 −0.484 −2.029 −1.095 −0.220 (0.445) (0.392) (0.165) (0.278) (0.648) (0.590) (0.752) (0.457) DT2_share −0.875 1.260 −0.534 −4.117 −0.499 −0.886 2.397 0.112 (0.855) (0.738) (0.349) (0.459) (1.347) (1.207) (1.524) (0.986) DT1_share 16.725 11.661 15.525 7.841 −20.177 15.491 10.100 15.962 (3.049) (2.576) (1.147) (1.691) (4.069) (4.042) (5.592) (3.473) Epil_share 9.578 11.917 4.069 4.048 −9.185 −13.955 −1.462 −0.165 (4.637) (3.910) (1.695) (2.815) (6.681) (5.882) (7.709) (4.754) Schi_share 23.551 28.298 37.453 39.029 21.821 39.502 1.307 29.586 (5.265) (4.248) (1.810) (3.082) (7.191) (6.259) (7.663) (5.136) Ln_ListLength −0.702 −0.631 −0.405 −0.658 −0.346 −0.489 −0.205 −0.623 (0.053) (0.047) (0.019) (0.033) (0.076) (0.069) (0.090) (0.052) GP_Age 0.032 0.029 0.033 0.035 0.032 0.033 0.033 0.033 (0.003) (0.003) (0.001) (0.002) (0.005) (0.004) (0.006) (0.003) GP_Sex −0.367 −0.512 −0.202 −0.234 −0.138 −0.108 −0.390 −0.317 (0.189) (0.166) (0.065) (0.118) (0.265) (0.235) (0.306) (0.175) GP Age Sex 0.010 0.013 0.006 0.007 0.009 0.004 0.010 0.008 (0.004) (0.003) (0.001) (0.002) (0.005) (0.005) (0.006) (0.004) GP_Specialist −1.148 −1.271 −1.145 −1.288 −1.119 −1.236 −1.189 −1.242 (0.035) (0.030) (0.012) (0.021) (0.050) (0.044) (0.056) (0.033) Pat_Sex −0.188 −0.090 −0.100 −0.082 −0.133 0.015 −0.163 0.040 (0.035) (0.028) (0.012) (0.020) (0.047) (0.041) (0.055) (0.032) Pat_BirthYear 0.007 0.007 0.012 0.007 0.003 0.007 0.012 0.195 (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.002) (0.036) Pat_Comorb 0.135 0.103 0.096 0.162 0.169 0.193 0.214 (0.027) (0.022) (0.013) (0.016) (0.034) (0.035) (0.038) Pat_Visits_win 0.042 0.046 0.049 0.041 0.035 0.049 0.046 0.057 (0.003) (0.003) (0.001) (0.002) (0.005) (0.004) (0.005) (0.004) Pat_Visits_dum 0.046 −0.208 −0.270 −0.327 −0.141 −0.157 −0.212 −1.019 (0.116) (0.087) (0.046) (0.083) (0.158) (0.134) (0.143) (0.333) Cons −12.977 −13.367 −25.141 −13.052 −7.795 −14.111 −26.068 −0.306 (1.863) (1.736) (0.738) (1.283) (2.508) (2.090) (3.519) (0.402) Sigma_u 0.718 0.784 0.773 0.747 0.755 0.809 0.922 0.662 (0.040) (0.032) (0.013) (0.024) (0.051) (0.043) (0.049) (0.042) No. obs 130,690 175,010 890,215 357,153 53,206 73,419 39,535 146,906 No. patients 27,157 37,110 186,415 74,473 11,292 15,403 8,316 30,212 Dependent variable: ‘SwitchOut’. Only patients living in cities with more than 50,000 inhabitants were included. The seven left columns are from sample 1, the far right column is from sample 2. For ‘Others’, ‘Pat_BirthYear’ was replaced with a dummy variable equal to 0 for patients born in 1940 and equal to 1 for patients born in 1970. Each patient was observed up to five times. Sigma_u denotes the estimated standard deviation of the random patient-specific constant terms. Stata 13, the xtlogit procedure, was used in the estimations. Estimates with two-sided p-values < 1% are in boldface Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 8 of 10 instance, a high share of DT1 patients increased the patients with the same diagnosis; for instance, ‘Arth_share’ switch-out for patients with arthritis (meaning, patients has a negative effect (−15.032) for patients with arthritis, with arthritis were more likely to switch-out if their GPs and ‘Asthm_share’ has a negative effect (−10.406) for had more patients with DT1). The cross share effect was patients with asthma. Again, this may be the result of GPs not generally symmetric as a high share of patients with informally specializing in certain types of patients with arthritis reduced the switch-out for patients with DT1. chronic diseases. It may also result from the GPs’ general For all GP and patient characteristics, the significant qualities such as organizational skills, communication coefficients had the same sign across all patient groups. abilities, or empathic attitudes. It has been suggested that Patients tended to switch less often from GPs who had such patterns may result from patients’ negative interac- long patient lists (‘Ln_ListLength’) or who were specialists tions with healthcare providers, so that, for instance, obese in general medicine (‘GP_Specialist’). For older, female patients search for “obese friendly” physicians [25]. GPs, patients tended to switch out more often (‘GP_Age’). Patients could also make use of informal conversations This effect was even stronger for male GPs, for which the (word-of-mouth) with family, friends, or colleagues that full effect of age is obtained by adding the coefficients of recommend one GP or another, which seems to have a ‘GP_age’ and the interaction between a GP’sage andsex greater effect on the choice of GP than public information (‘GP_Age*GP_Sex’). disclosure [20]. The relationship between the GP and Patients born more recently (i.e., lower ‘Pat_BirthYear’) patient could also be a factor in patient choice, since or who had more comorbidities (‘Pat_Comorb’) tended chronic patients spend more time in primary care and to switch GPs more often. The 1% of patients who most would change their GP if they were not satisfied [3, 4]. We frequently used primary care (i.e., ‘Pat_Visits_dum’ =1) can assume that GPs who have high numbers of patients tended to switch less often than patients who had fewer with a particular disease might have a particular practice visits. However, among the remaining 99% of patients, style, which also attracts these patients, but these mecha- those with a higher number of primary care visits nisms may be complex, for instance for patients with (‘Pat_visits_win’)tendedtoswitchmoreoften. schizophrenia. In Table 5, the only exception from the The patient-specific effects are assumed to be normally general pattern is for patients with schizophrenia, for distributed, with a zero mean and an estimated standard which the effect of ‘Schi_share’ is insignificant. However, deviation, Sigma_u. For patients with arthritis, the value all other patient groups tend to disenroll more from GPs of Sigma_u can be interpreted as the difference in log- with high shares of patients with schizophrenia, poten- odds between a patient who has a patient-specific inter- tially suggesting that these GPs are less popular in general, cept one standard deviation from the mean (0.718) and and this may perhaps counter the “own share effect” a patient with an intercept equal to the mean value among patients with schizophrenia. (zero). This is about four times the numerical value of We find that all or most patient groups tend to disen- the coefficient for patient sex, and it corresponds to a roll less from GPs who have high shares of patients with difference in odds ratio equal to 2.050. In all patient arthritis, depression, and asthma. We assume that this groups, the estimated value for Sigma_u indicates that disenrollment pattern happens due to qualities of GPs the unobserved patient characteristics have a comparably that attract most patients, such as good communication large influence on disenrollment. and care coordination skills. For chronic patients who are intensive users of primary care it is important to find Discussion a GP that fits their needs, so they might change until Our data indicate that patients with chronic diseases are they find the right match. Patients in the comparison not allocated to GPs by chance alone (Fig. 2). One ex- group have, per se, no obvious reason to prefer GPs who planation could be that some GPs informally specialize, specialize in any chronic disease, but it is likely they have for example in DT2, and thus are able to establish and preferences regarding GP qualities. Thus, our finding maintain a “stock” of such patients. In so doing, the that in some cases the preferences of the comparison patient comorbidity shown in Table 3 would imply a group and of the patients with chronic diseases align tendency for these GPs to also have relatively higher suggests that GPs’ shares of chronic patients reveals shares of patients with arthritis and asthma. Moreover, information about these GPs’ general qualities. patients with chronic diseases tend to have comorbidities, A puzzling finding is that all or most patient groups contributing to their GPs having shares of patients with tend to disenroll more from GPs who have high shares of different diagnoses. This could partly explain why the patients with DT1 and schizophrenia. According to proportions of chronic disease types are all positively Norwegian guidelines, these two patient groups’ follow-up correlated, as shown in Table 4. happens in secondary care, in contrast to our other patient The coefficients in Table 5 suggest that chronic patients groups. Patients who receive follow-up in secondary care disenroll less often from GPs who have a high share of could perhaps be more indifferent to which GP they visit Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 9 of 10 for other acute illnesses. If so, they may be satisfied with (‘DT1_share’ and ‘DT2_share’), and the results for sub- GPs who have a practice style favoring patients who can samples defined for patients with DT1 and DT2. Second, be treated expediently over patients who need long-term our data did not include potentially relevant patient follow-up. With this interpretation, the high disenrollment variables such as cultural background, native language, among patients with schizophrenia (Table 2) can be income, educational background, or marital status. Disease interpreted not necessarily as a search for a GP who is severity and proper control of symptoms could also in- well-suitedforhandlingissues relatedtoschizophrenia fluence disenrollment behavior. To an extent, our ran- but perhaps as an expression of other, shorter-term dom effect logistic regressions can account for time- considerations. invariant patient variables, but future studies should GP specialization in general medicine has a negative consider including more variables in order to assess relationship with disenrollment, suggesting that patients their influence. Additional information about the GPs, prefer to stay with specialized GPs. List length also has a such as cultural background, length of time in practice, negative relationship with disenrollment for all patient and professional interests would also have been of groups, except for patients with schizophrenia. Previous interest. Third, the age distribution differs between our studies have found that non-chronic patients stay with selected comparison group, sample 2, and our main GPs with shorter patient lists, meaning that they value sample of interest, sample 1. Sample 2’sage distribution accessibility [10–12], in contrast to chronic patients who also differs from the age distribution across all groups value long patient lists, which is associated with higher in the full population without our specified chronic disease detection [13]. GP’s age is positively related with diseases.Thismeansthatthe estimatesfor sample 2in disenrollment for all patient groups, suggesting that pa- Tables 2 and 5 are likely to be biased, if interpreted as tients in general may prefer younger GPs. This effect of estimates for the full population. We believe that the age is supported by earlier findings [12]. For patients qualitative aspects of these results would not be very with arthritis, asthma, depression or DT2, this tendency different in the full population, but this is of course a is stronger for male than female GPs, perhaps because conjecture. Future register-based studies should consider there are fewer women among older GPs than among obtaining a comparison group with similar age distribu- younger GPs. In most patient groups, disenrollment was tion as the sample of main interest, for instance by not significantly associated with GP sex, except patients drawing patients randomly from the entire population. with asthma and depression, who tend to less often The data sets used in our logistic regressions were re- disenroll from male GPs. stricted with respect to municipality size. In smaller In all groups of patients with chronic disease, disen- municipalities, patient options for disenrollment will be rollment increased with the number of comorbidities. more limited by the fact that there are fewer local GPs This is consistent with the discussion above, given that to choose from. It is likely that including patients irre- management of patients with comorbidities is challen- spective of municipality size would yield estimated ging for primary care providers [27]. Our selection of effects less pronounced than those reported here – that patient groups was not, however, designed to investigate is, compared to the full population, our result are likely the effect of comorbidities in particular. Future studies to be biased away from zero. We also excluded observa- should consider including other diagnoses, such as cardio- tions where observed disenrollment seemed to be due vascular disease and cancer. A higher number of visits to to causes other than patients’ preferences for GPs. primary care also tended to increase disenrollment, but Patients and GPs who move, or GPs who retire or die, the negative coefficients for the dummy variable, identi- are likely to have demographic characteristics (e.g., age) fying patients who had more than 23 visits in a six that differ systematically from the distributions in the month period, may indicate that the relationship full patient and GP populations. It is more difficult to between disenrollment and the number of visits is not predict how including these observations would have linear. Younger patients generally disenroll more often influenced our results, but it would at least have com- and, except for patients with epilepsy and other patients plicated the interpretations. (sample 2), male patients disenroll less often. This study has three main imitations: first, although Conclusions the majority of the numerical data seemed reliable, we The following conclusions can be drawn from our found that as many as 77.8% of patients with DT1 were findings: 1) patients with chronic diseases are not allo- also registered as having DT2. Such “double diabetes” cated to GPs only by chance; 2) chronic patients that cases are not uncommon [28, 29], but it is likely that use primary care intensively disenroll less often from most of the cases in our data are due to diagnostic un- GPs who have a high share of patients with the same certainty or registration errors. This may affect both the diagnosis; and 3) most patient groups tend to remain results related to the share of patients with diabetes with GPs with a greater share of arthritis, asthma, and Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 10 of 10 depression patients, which can indicate better quality 5. Saultz JW, Lochner J. Interpersonal continuity of care and care outcomes: a critical review. Ann Fam Med. 2005;3(2):159–66. care for these and other patient groups. These conclu- 6. Safran DG, et al. Switching doctors: predictors of voluntary disenrollment sions are distinct from the findings in the literature. from a primary physician's practice. J Fam Pract. 2001;50(2):130–6. To investigate this further, more objective quality mea- 7. Nagraj S, et al. Changing practice as a quality indicator for primary care: analysis of data on voluntary disenrollment from the English GP Patient surements should be obtained, such as adherence to Survey. BMC Fam Pract. 2013;14:89. treatment guidelines, surveillance of treatment outcomes 8. Kasteler J, et al. Issues underlying prevalence of “doctor-shopping” behavior. for chronic patients, and user satisfaction in general. If J Health Soc Behav. 1976;17(4):329–39. 9. Gray PG, Cartwright A. Choosing and changing doctors. Lancet. 1953; objective quality differences are found, further assess- 265(6799):1308–9. ments could be warranted, for instance, whether the 10. Marquis MS, Davies AR, Ware Jr JE. Patient satisfaction and change in current reimbursement system has an appropriate medical care provider: a longitudinal study. Med Care. 1983;21(8):821–9. 11. Rubin G, et al. Preferences for access to the GP: a discrete choice balance between capitation and fee-for service, or experiment. Br J Gen Pract. 2006;56(531):743–8. whether capitation should be risk-adjusted based on 12. Campbell JL, et al. Accessing primary care: a simulated patient study. Br J shares of patient types. Gen Pract. 2013;63(608):e71–6. 13. Campbell SM, et al. Identifying predictors of high quality care in English general practice: observational study. BMJ. 2001;323(7316):784–7. Abbreviations 14. Heje HN, et al. Doctor and practice characteristics associated with DT1: Type 1 diabetes; DT2: Type 2 diabetes; GP: General practitioner; differences in patient evaluations of general practice. BMC Health Serv Res. KUHR: Control and payment of reimbursements to health service providers 2007;7:46. (Kontroll og Utbetaling av HelseRefusjon) 15. Anwar MS, et al. Chronic disease detection and access: does access improve detection, or does detection make access more difficult? Br J Gen Pract. Acknowledgements 2012;62(598):e337–43. We wish to thank Tor Iversen for many valuable comments. The Norwegian 16. Soljak M, et al. Variations in cardiovascular disease under-diagnosis in Directorate of Health provided data. England: national cross-sectional spatial analysis. BMC Cardiovasc Disord. 2011;11:12. Funding 17. Norbury M, et al. Time to care: tackling health inequalities through primary The Research Council of Norway, project no. 204677/H10, provided funding. care. Fam Pract. 2011;28(1):1–3. 18. Mercer SW, Watt GC. The inverse care law: clinical primary care encounters Availability of data and materials in deprived and affluent areas of Scotland. Ann Fam Med. 2007;5(6):503–10. The datasets analyzed during the current study were obtained from the 19. Santos R, Gravelle H, and Propper C. Does quality affect patients’ choice of Norwegian Directorate of Health. Due to legal restrictions we are prevented doctor? Evidence from England. Econ J. 2016. doi:10.1111/ecoj.12282. from making the data publicly available or otherwise sharing individual level 20. Marshall MN, Shekelle PG, Leatherman S, Brook RH. The public release of data. For access to such data for research purposes, please contact the performance data: what do we expect to gain? A review of the evidence. Norwegian Directorate of Health directly. Jama. 2000;283(14):1866–74. 21. Marshall M, McLoughlin V. How do patients use information on health Authors’ contributions providers? BMJ. 2010;341:c5272. AM conceived the study, prepared the data and performed the statistical 22. Lurås H. The association between patient shortage and patient satisfaction calculations. Both AM and KRW participated in the study design, with general practitioners. Scand J Prim Health Care. 2007;25(3):133–9. interpretation of results, and writing of the manuscript. The final manuscript 23. Williams SJ, Calnan M. Key determinants of consumer satisfaction with has been read and approved by both AM and KRW. general practice. Fam Pract. 1991;8(3):237–42. 24. Sitzia J, Wood N. Patient satisfaction: a review of issues and concepts. Soc Competing interests Sci Med. 1997;45(12):1829–43. There author declare that there have no competing interests. 25. Gudzune KA, Bleich SN, Richards TM, Weiner JP, Hodges K, Clark JM. Doctor shopping by overweight and obese patients is associated with increased Consent for publication healthcare utilization. Obesity. 2013;21(7):1328–34. The patient data were obtained from public registers and patient consent for 26. Kristjansson E, et al. Predictors of relational continuity in primary care: publication was not required. patient, provider and practice factors. BMC Fam Pract. 2013;14:72. 27. Wallace E, et al. Managing patients with multimorbidity in primary care. Ethics approval and consent to participate BMJ. 2015;350:h176. The patient data were obtained from public registers and patient consent to 28. Pozzilli P, et al. Obesity, autoimmunity, and double diabetes in youth. participate was not required. The project has been approved by the Regional Diabetes Care. 2011;34 Suppl 2:S166–70. Committees for Medical and Health Research Ethics (REC South East), project 29. Cleland SJ, et al. Insulin resistance in type 1 diabetes: what is ‘double no. 2011/1708. diabetes’ and what are the risks? Diabetologia. 2013;56(7):1462–70. Received: 21 June 2016 Accepted: 2 December 2016 References 1. Starfield B. Commentary on regular primary care lowers hospitalisation risk and mortality in seniors with chronic respiratory disease. J Gen Intern Med. 2010;25(8):758–9. 2. Harris MF, Zwar NA. Care of patients with chronic disease: the challenge for general practice. Med J Aust. 2007;187(2):104–7. 3. Starfield B, Lemke KW, Herbert R, Pavlovich WD, Anderson G. Comorbidity and the use of primary care and specialist care in the elderly. Ann Fam Med. 2005;3(3):215–22. 4. Kaplan SH, Greenfield S, Ware Jr JE. Assessing the effects of physician- patient interactions on the outcomes of chronic disease. Med Care. 1989; 27(3 Suppl):S110–27. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Family Practice Springer Journals

Disenrollment from general practitioners among chronic patients: a register-based longitudinal study of Norwegian claims data

BMC Family Practice , Volume 17 (1) – Dec 15, 2016

Loading next page...
 
/lp/springer-journals/disenrollment-from-general-practitioners-among-chronic-patients-a-aKbKSjYOzb

References (29)

Publisher
Springer Journals
Copyright
Copyright © 2016 by The Author(s).
Subject
Medicine & Public Health; General Practice / Family Medicine; Primary Care Medicine
eISSN
1471-2296
DOI
10.1186/s12875-016-0571-3
pmid
27978811
Publisher site
See Article on Publisher Site

Abstract

Background: Norwegian general practitioners (GPs) consult on a variety of conditions with a mix of patient types. Patients with chronic diseases benefit from appropriate continuity of care and generally visit their GPs more often than the average patient. Our aim was to study disenrollment patterns among patients with chronic diseases in Norway, because such patterns could indicate otherwise unobserved GP quality. For instance, higher quality GPs could have both a greater share of patients with chronic diseases and lower disenrollment rates. Methods: Data on 384,947 chronic patients and 3,974 GPs for the years 2009–2011 were obtained from national registers, including patient and GP characteristics, disenrollment data, and patient list composition. The birth cohorts from 1940 and 1970 (146,906 patients) were included for comparison. Patient and GP characteristics, comorbidity, and patient list composition were analyzed using descriptive statistics. Patients’ voluntary disenrollment was analyzed using logistic regression models. Results: The GPs’ proportion of patients with a given chronic disease varied more than expected when the allocation was purely random. The proportions of patients with different chronic diseases were positively correlated, partly due to comorbidity. Patients tended to have lower disenrollment rates from GPs who had higher shares of patients with the same chronic disease. Disenrollment rates were generally lower from GPs with higher shares of patients with arthritis or depression, and higher from GPs who had higher shares of patients with diabetes type 1 and schizophrenia. This was the same in the comparison group. Conclusion: Patients with a chronic disease appeared to prefer GPs who have higher shares of patients with the same disease. High shares of patients with some diseases were also negatively associated with disenrollment for all patient groups, while other diseases were positively associated. These findings may reflect the GPs’ general quality, but could alternatively result from the GPs’ specialization in particular diseases. The supportive findings for the comparison group make it more plausible that high shares of chronic patients could indicate GP quality. Keywords: Chronic patients, Switching, Primary health care, Schizophrenia, Epilepsy, Diabetes type 1, Diabetes type 2, Asthma, Arthritis, Depression * Correspondence: anastasiya.mokienko@gmail.com; anastasiya.mokienko@gmail.com Department of Health Management and Health Economics, University of Oslo, P.O. Box 1089, Blindern, Oslo 0318, Norway © The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 2 of 10 Background If patients switch between GPs until their demands are The quality of care for people with chronic diseases met, we would expect these patients to be dispropor- often relies on appropriate primary care. Some such tionally distributed across GPs. Similar trends could be patients may need continuous, long-term follow-up and expected if the GPs intentionally specialize, formally or motivation in order to maintain a favorable lifestyle. not, in a given patient group. However, neither of these Others, who experience a condition associated with mechanisms has obvious implications for the provider social stigma, may need time to develop trust in their choices made by other groups of patients. For example, care providers. Early detection of the chronic disease a GP who is popular among patients with diabetes type and its subsequent routine monitoring is also very 2 (DT2) may also be popular among patients with important to save patients from acute hospitalization depression, whereas patients without chronic diseases and complications from the disease [1]. Comorbidity is a may be indifferent to this GP’s motivational skills. Older good reason for primary care providers to be better able patients and patients with chronic diseases have generally to manage chronic diseases [2, 3]. higher care continuity, whereas patients with lower care Previous studies have found that long-term physician- continuity are those living in rural areas, employed, with patient relationships are beneficial for patients [4, 5] and higher education, or with poorer mental health [26]. that patients disenroll from their general practitioner Our aim is to investigate patterns of chronic patient (GP) when they are not satisfied with their GP-patient disenrollment. This type of study is required because relationship [6–10]. Patients may also disenroll from their there are no published indicators of GP quality, and GP if they perceive insufficient quality of care. Accessibility therefore these indicators need to be identified through factors, such as adequate time for consultations [11] and patient actions (such as disenrollment). Moreover, spe- availability of appointments [12] are predictors of good cialized patient choice patterns might suggest an extra quality. Booking intervals for consultations and duration argument for using more fee-for-service reimbursement of the consultations themselves are correlated with good or risk-adjusted capitation for GPs in order to compen- management of chronic diseases; the effect was greater for sate for varying expected workloads depending on their patients with asthma than for those with diabetes or patient list composition. Primary care in Norway is pub- angina, possibly because primary care providers deal licly funded with a capitation and fee-for-service system, more with asthma than diabetes or angina [13]. and patients have to consult their GPs in order to see a When it comes to accessibility, earlier research shows specialist. Each individual GP has a patient list and can that longer patient lists are associated with negative decide the maximum number of patients that can be evaluations of accessibility and that the GP's age has a enrolled on their list. Patients can switch between available negative association with the evaluation of all aspects, GPs up to three times a year, according to their own except accessibility [14]. Longer patient lists are also asso- preference. ciated with better illness detection [15], which may sug- gest that practices detecting a higher number of chronic Methods conditions have greater demand from patients due to their Data sources and study populations systematic chronic disease management [15–18]. This is a retrospective study using data from two national A strong connection between patient choice and registers in Norway, administrated by the Norwegian higher quality of practice, as measured by studying the Directorate of Health, from 2009–2011. Our GP data publicly available data on practice performance, has been were obtained from the national register of regular reported [19]. A review study found that patients were GPs, which covers the entire GP population, and weakly influenced by publicly available information merged with patient data using the GPs’ IDs. Our patient about provider quality [20]. On the provider side, only data were based on claims data obtained from the KUHR hospitals seemed to improve quality as a response to registry (Kontroll og Utbetaling av HelseRefusjon), which quality indicators being made publicly available [21]. For covers the entire Norwegian patient population. This GPs, patient shortage has been found to correlate with registry records claims data continuously but for our ana- patient dissatisfaction, the GP’s communication skills, lysis, the sample period 2009–2011 was divided into six and other GP characteristics [22–24]. semiannual intervals. The individual level data included Interaction between chronically ill patients and their patient characteristics, their consumption of primary care, GPs has not been given specific attention in previous and the GP with which they were enrolled. literature, but a previous study of obese patients may Two samples of patients were selected among patients contain clues for generalizable results: reportedly, obese who visited a GP at least once from 2009–2011. Most of patients avoided physicians they perceived as sources of our analysis is based on sample 1, which consisted of pa- stigma and searched for providers who were “obese tients registered with one or more of the following seven friendly” [25]. diagnoses at least once during the period 2006–2011: Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 3 of 10 DT1, DT2, asthma, arthritis, schizophrenia, depression, Measures and epilepsy. These patient groups were chosen because Our main outcome variable, ‘SwitchOut’, measured they are known to vary substantially both in the number whether a patient disenrolled from a GP from one semi- of patients in the population, and in the utilization of annual period to the subsequent period. Definitions of primary care services. For instance, patients with DT2 independent variables are summarized in Table 1. Infor- constitute almost 5% of the population and receive most mation about the GPs’ age, sex, specialization, and list of their health care from their GP, while patients with length, and patients’ sex, birth year, and number of visits schizophrenia are fewer and receive more specialist care was obtained directly from the data registries. The vari- in a hospital setting. able ‘Pat_comorb’ was given the value 0 for patients in Our analysis also included a comparison group, sample 2. sample 2, while for each patient in sample 1 we counted This group consisted initially of the entire birth year the number of registered diseases (1–7) and subtracted 1 cohorts from 1940 and 1970, but we excluded patients from this number. This yielded a variable with a range be- already included in sample 1. Obviously this selection tween 0 and 6. The variables ‘Diab2_share’ and ‘Epil_share’ yielded an age distribution different from that in sample measure a GP’s share of patients with the respective 1, but the selection of one elderly and one younger chronic disease, but with a slight adjustment: if shares birth year cohort should provide a good basis for were calculated straightforwardly, they could potentially comparison. be influenced by the health status of a single patient, Initially, the two samples combined contained 988,483 because some chronic diseases are relatively rare and patients (Fig. 1). We excluded 34,189 cases where the some GPs had fewer patients (shorter lists). To illus- disenrollment was likely to be due to causes not relevant trate, consider a GP who has 100 patients, of which one for our purpose; that is, when patients moved to another has epilepsy. If we take the perspective of the GP, the municipality, or when a GP moved, retired, or died. For share of patients with epilepsy is slightly above average the logistic regressions, we excluded patients living in (Table 1). However, this measure is of little relevance if municipalities with less than 50,000 inhabitants in order we take the perspective of the patient with epilepsy: the to focus on patients who could choose from several GPs. GP has no other patients with epilepsy. To avoid inter- This left us with 316,636 patients in sample 1 and 32,311 pretational ambiguity, we chose to take the patients’ patients in sample 2 (348,947 in total). Finally, we perspective. For each patient-GP pair, we excluded the excluded patients with irregular medical records, mainly patient from the calculation of the GP’s share. Thus, the missing birth year or sex, yielding 313,659 patients in sam- share variables mostly showed the variation between GPs ple 1 and 30,212 patients in Sample 2 (343,871 in total). but also some variation within a GP practice. Fig. 1 Flow diagram of sample selection Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 4 of 10 Table 1 Variable definitions and descriptive statistics on the patient level Variable Definition Sample 1 (N = 313,659) Sample 2 (N = 30,212) Median Mean St.dev Median Mean St.dev DT1_share The share of a GP’s patients with diabetes type 1 0.005 0.007 0.006 0.005 0.006 0.005 DT2_share The share of a GP’s patients with diabetes type 2 0.042 0.046 0.027 0.036 0.040 0.021 Arth_share The share of a GP’s patients with arthritis 0.014 0.016 0.010 0.013 0.015 0.009 Asthm_share The share of a GP’s patients with asthma 0.020 0.023 0.015 0.018 0.020 0.013 Depr_share The share of a GP’s patients with depression 0.107 0.112 0.042 0.094 0.100 0.038 Schi_share The share of a GP’s patients with schizophrenia 0.004 0.005 0.003 0.004 0.005 0.003 Epil_share The share of a GP’s patients with epilepsy 0.008 0.009 0.004 0.008 0.008 0.004 ListLength The number of patients on a GP’s list 1423 1444.0 367.8 1439 1453.4 367.8 Ln_ListLength The natural logarithm of Listlength 7.261 7.240 0.277 7.272 7.248 0.270 GP_Age The GP’s age 52 50.358 9.120 51 49.744 8.989 GP_Sex =1 if the GP is male, =0 otherwise 1 0.706 0.455 1 0.673 0.469 GP_age · GP_Sex The product of GP_Age and GP_Sex 48 36.473 24.718 45 34.266 25.016 GP_Specialist =1 if the GP has a specialist degree 1 0.707 0.455 1 0.702 0.457 in general medicine; =0 otherwise Pat_Sex =1 if the patient is male; =0 otherwise 0 0.426 0.494 0 0.494 0.500 Pat_BirthYear The patient’s year of birth 1959 1958.6 19.1 1970 1961.5 13.5 Pat_Comorb Sample 1: No. of chronic diseases minus one. 0 0.148 0.405 - Sample 2: Not defined Pat_Visits The patient’s number of visits to primary care 3 4.662 5.268 1 2.227 3.369 th Pat_Visits_win Winsorized Pat_Visits at 99 percentile (max = 23) 3 4.570 4.626 1 2.205 3.107 Pat_Visits_dum =1 if Pat_Visit >23, =0 otherwise 0 0.10 0.98 0 0.002 0.047 Municipalities over 50 000. First half of 2009 In order to avoid highly influential outliers, we trans- sub-samples partly overlapped due to comorbidity. For formed two variables. The distribution of GPs’ list length each sub-sample, the shares of patients with 1 of the was skewed so we transformed the variable using the other six diseases were calculated. natural logarithm. The distribution of patients’ number We then used logistic regressions to model patients’ of visits to primary care was also skewed, and for this disenrollment from their GP. The modeling was per- th variable, we winsorized the distribution at the 99 per- formed for each patient category separately: on the sub- centile (23 visits per period) and included a dummy vari- samples from sample 1, as defined above, and sample 2. able for observations that exceeded this limit. Because the dependent variable (SwitchOut) was based on observations from two consecutive periods, we had Statistical analyses up to five effective observations for each patient. For the We inspected the data numerically and graphically at independent variables, we used observations from the both the patient and GP levels. This included graphs first five periods. The set of independent variables intended to reveal whether the distribution of chronic included those from Table 1, and an interaction term patients seemed disproportionate across GPs. On the GP between GPs’ age and sex. We incorporated the longitu- level, the mean proportion of patients with DT2 was dinal data structure by including patient-specific effects 4.5% in the first half of 2009. If patients were allocated (intercepts) in the models. Patient-specific effects can by pure chance, a randomly selected GP’s share of account for unobserved factors, such as ethnicity or patients with DT2 would have the expected value of educational background, as long as these factors remain about 4.5%, and be approximately normally distributed constant throughout the sample period. The models were for a sufficiently long patient list (>60 patients). For data estimated using xtlogit in Stata 13, under the standard at the GP level, we calculated Spearman’s correlation co- assumptions that the patient-specific effects were nor- efficients for the various GP-related variables, including mally distributed and did not correlate with the inde- the shares of patients with different diagnoses, the GP’s pendent variables. Fixed effect models, which allow the age and sex. We defined sub-samples of patients from patient-specific effects to be non-normally distributed or sample 1 based on the seven chronic diseases. These correlated with the independent variables, were also Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 5 of 10 considered. However, in fixed effect models the time- Table 2 Share of patients who voluntarily disenrolled from their st nd 1 GPs, between the 1 and 2 halves of 2009. invariant patient variables for sex and birth year would, by construction, be excluded from the estimations. Sample Sub set N % Sample 1 Full sample 313,659 4.52 Results DT1 11,292 4.99 Descriptive statistics DT2 74,473 3.75 According to Table 1 and Fig. 2, the proportion of patients Schizo 8,316 6.29 with DT2 varied substantially among GPs. If these patients Depr 186,415 5.00 had been allocated purely by chance, about 95% of the Arthr 27,157 4.00 proportions would lie between the red curves in Fig. 2, but this was not the case. In fact, only 46.5% of the pro- Asthm 37,110 4.16 portions were positioned within the red curves. For the Epil 15,403 4.86 other diagnosis groups, the corresponding patient shares Sample 2 Full sample 30,212 3.76 also seemed disproportionally distributed. Municipalities over 50 000 Overall, 4.5% of chronic patients disenrolled from their GP from one period to the next, but the share varied from 3.7% among patients with DT2 to 6.2% among was the least frequent (N = 21,368). In the sub-sample of patients with schizophrenia (Table 2). Among patients in patients with depression (third column from the left), sample 2, the share that disenrolled was 3.7%. 1.3% also suffered from schizophrenia. Among patients Descriptive statistics for the independent variables with schizophrenia (rightmost column), 28.7% also suf- used in the logistic regressions are reported in Table 1, fered from depression. A substantial number of patients separately for samples 1 and 2. Due to the sample selec- were recorded with both DT1 and DT2, likely due to tion procedure, the average GP characteristics differ some- registration errors or diagnostic uncertainty. what from those obtained for the full GP population, We calculated Spearman’s rank correlation coefficients where 66% were men, the average age was 48 years, and for the GP proportion of patients with a given chronic the average patient list length was 1200 (N = 3940). disease and other patient proportions and GP charac- The distribution of the variable ‘ListLength’ appeared teristics, as shown in Table 4. The correlation coeffi- continuous but was somewhat skewed to the right. The cient of ‘Asthm_share’ and ‘DT1_share’ was 0.648, distribution of ‘Pat_visits’ was markedly right-skewed, and indicating that GPs with a high proportion of patients the distribution’s tail was rather scattered: for sample 2, with asthma also tended to have a high proportion of th the 75th, 95 , and 99th percentiles were 6, 14, and 23, patients with DT1. All variables related to the GPs’ pro- respectively, but the maximum value was as high as 219. portions of patients were significantly different from Table 3 presents the sizes of the sub-samples defined for zero. The proportion of patients with chronic diseases the seven chronic diseases. The most frequent of the dis- were all positively correlated, and negatively correlated eases was depression (N = 488,686), while schizophrenia with the proportion of other patients (‘Other_Share’). ‘Other_Share’ was negatively correlated with ‘GP_Age’ and ‘GP_Sex’, indicating that older GPs and male GPs tended to have fewer patients without our seven chronic diseases. Logistic regression analysis Table 5 shows the estimated parameters of the logistic regressions where ‘SwitchOut’ is the dependent variable, the independent variables are those listed in Table 1, and Sigma_u denotes the standard deviation of the patient- specific intercepts. The first seven columns show results based on sample 1 according to patient diagnosis group; the last column is based on sample 2. In logistic regres- sions, the coefficients can be used to compare the differ- ence in log-odds ratios between groups, so that a patient Fig. 2 Scatter plot of GP proportion of DT2 patients and patient list sex coefficient of −0.188 (arthritis patients) represents the length. Legend: Y-axis percent of DT2 patients, X-axis patient-list difference in log-odds ratios between male and female length. GP level, data for the first quarter of year 2009, N = 3,965, patients. The corresponding difference in odds ratios is mean proportion of DT2 patients = 0.045, patient-list lengths of >60 obtained by taking the anti-log, exp(−0.188) = 0.829. Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 6 of 10 Table 3 Percent of patients with a chronic disease (column) that have another chronic disease (row) Arthritis Asthma Depression DT2 DT1 Epilepsy Schizophrenia Arthritis 4.4 2.7 3.9 4.0 2.0 1.3 Asthma 6.1 4.5 6.7 5.8 3.6 6.0 Depression 14.5 17.6 13.5 15.2 15.8 28.7 Diabetes type 2 10.0 12.4 6.4 77.8 5.7 12.0 Diabetes type 1 1.6 1.7 1.1 12.0 1.3 1.9 Epilepsy 1.0 1.3 1.5 1.1 1.6 3.1 Schizophrenia 0.3 1.0 1.3 1.1 1.1 1.4 N 90,095 124,776 488,686 232,383 35,887 46,145 21,368 First half year of 2009. Patient level data. Sample 1 without restrictions (neither on municipality size, data irregularity or moving). N is the number of patients with the chronic disease The statistical inference for this type of model is based disenrollment from GPs with relatively high shares of pa- on large-sample theory and coefficient estimates are tients with arthritis. For ‘Asthm_share’ and ‘Depr_share’, approximately normally distributed. Thus, to simplify all of the significant coefficients were also negative. In the presentation, we do not report p-values as they can contrast, for ‘DT1_share’, ‘Epil_share’ and ‘Schi_share’, be derivedfromthe estimatedstandarderrors. almost all significant effects were positive. Some of the estimated effects of the patient share vari- We can distinguish two main effects. First, the “own ables were relatively robust across patient groups. For share effect,” namely, all patient groups tended to remain ‘Arth_share’, all coefficients were significantly negative, with GPs who had a high share of patients with the same implying that all patient groups tended to have lower diagnosis. Second, the “cross share effect,” where, for Table 4 GP characteristics. Spearman’s correlation coefficients with two-sided p-values. Arth_ Asthm_ Depr_ DT1_ DT2_ Epil_ Schi_ Other_ GP_ GP_ List share share share share share share share share age sex Length Asthm_share 0.488 0.000 Depr_share 0.195 0.264 0.000 0.000 DT1_share 0.519 0.648 0.221 0.000 0.000 0.000 DT2_share 0.232 0.310 0.121 0.332 0.000 0.000 0.000 0.000 Epil_share 0.270 0.298 0.205 0.335 0.177 0.000 0.000 0.000 0.000 0.000 Schi_share 0.045 0.175 0.227 0.135 0.183 0.162 0.005 0.000 0.000 0.000 0.000 0.000 Other_share −0.562 −0.683 −0.762 −0.712 −0.362 −0.406 −0.285 0.000 0.000 0.000 0.000 0.000 0.000 0.000 GP_Age 0.203 0.137 0.064 0.213 −0.047 0.091 −0.028 −0.174 0.000 0.000 0.000 0.000 0.003 0.000 0.077 0.000 GP_Sex 0.181 0.293 0.077 0.318 0.101 0.205 0.135 −0.265 0.249 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ListLength −0.020 −0.069 0.041 −0.032 −0.145 −0.040 −0.033 0.035 0.166 0.172 0.205 0.000 0.010 0.046 0.000 0.011 0.038 0.026 0.000 0.000 GP_Specialist 0.008 0.017 0.030 0.037 −0.133 0.067 −0.003 −0.018 0.365 0.098 0.226 0.618 0.275 0.063 0.020 0.000 0.000 0.860 0.250 0.000 0.000 0.000 GP level data for first quarter of 2009, N = 3974. Correlation coefficients with two-sided p-values less than 1% are in boldface Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 7 of 10 Table 5 Logistic regression for patients’ voluntary disenrollment from GPs, separate for patient groups. Estimated parameters (standard errors) Arthritis Asthma Depression Diabetes type 2 Diabetes type 1 Epilepsy Schizophrenia Others Arth_share −15.032 −10.550 −16.792 −9.506 −16.905 −16.495 −20.113 −15.310 (1.611) (1.597) (0.815) (1.194) (3.116) (2.836) (3.925) (2.185) Asthm_share −4.381 −10.406 −2.117 1.883 −1.624 −0.188 −3.895 0.093 (1.598) (1.309) (0.636) (0.934) (2.494) (2.262) (2.922) (1.799) Depr_share −1.915 −2.343 −5.377 −2.781 −0.484 −2.029 −1.095 −0.220 (0.445) (0.392) (0.165) (0.278) (0.648) (0.590) (0.752) (0.457) DT2_share −0.875 1.260 −0.534 −4.117 −0.499 −0.886 2.397 0.112 (0.855) (0.738) (0.349) (0.459) (1.347) (1.207) (1.524) (0.986) DT1_share 16.725 11.661 15.525 7.841 −20.177 15.491 10.100 15.962 (3.049) (2.576) (1.147) (1.691) (4.069) (4.042) (5.592) (3.473) Epil_share 9.578 11.917 4.069 4.048 −9.185 −13.955 −1.462 −0.165 (4.637) (3.910) (1.695) (2.815) (6.681) (5.882) (7.709) (4.754) Schi_share 23.551 28.298 37.453 39.029 21.821 39.502 1.307 29.586 (5.265) (4.248) (1.810) (3.082) (7.191) (6.259) (7.663) (5.136) Ln_ListLength −0.702 −0.631 −0.405 −0.658 −0.346 −0.489 −0.205 −0.623 (0.053) (0.047) (0.019) (0.033) (0.076) (0.069) (0.090) (0.052) GP_Age 0.032 0.029 0.033 0.035 0.032 0.033 0.033 0.033 (0.003) (0.003) (0.001) (0.002) (0.005) (0.004) (0.006) (0.003) GP_Sex −0.367 −0.512 −0.202 −0.234 −0.138 −0.108 −0.390 −0.317 (0.189) (0.166) (0.065) (0.118) (0.265) (0.235) (0.306) (0.175) GP Age Sex 0.010 0.013 0.006 0.007 0.009 0.004 0.010 0.008 (0.004) (0.003) (0.001) (0.002) (0.005) (0.005) (0.006) (0.004) GP_Specialist −1.148 −1.271 −1.145 −1.288 −1.119 −1.236 −1.189 −1.242 (0.035) (0.030) (0.012) (0.021) (0.050) (0.044) (0.056) (0.033) Pat_Sex −0.188 −0.090 −0.100 −0.082 −0.133 0.015 −0.163 0.040 (0.035) (0.028) (0.012) (0.020) (0.047) (0.041) (0.055) (0.032) Pat_BirthYear 0.007 0.007 0.012 0.007 0.003 0.007 0.012 0.195 (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.002) (0.036) Pat_Comorb 0.135 0.103 0.096 0.162 0.169 0.193 0.214 (0.027) (0.022) (0.013) (0.016) (0.034) (0.035) (0.038) Pat_Visits_win 0.042 0.046 0.049 0.041 0.035 0.049 0.046 0.057 (0.003) (0.003) (0.001) (0.002) (0.005) (0.004) (0.005) (0.004) Pat_Visits_dum 0.046 −0.208 −0.270 −0.327 −0.141 −0.157 −0.212 −1.019 (0.116) (0.087) (0.046) (0.083) (0.158) (0.134) (0.143) (0.333) Cons −12.977 −13.367 −25.141 −13.052 −7.795 −14.111 −26.068 −0.306 (1.863) (1.736) (0.738) (1.283) (2.508) (2.090) (3.519) (0.402) Sigma_u 0.718 0.784 0.773 0.747 0.755 0.809 0.922 0.662 (0.040) (0.032) (0.013) (0.024) (0.051) (0.043) (0.049) (0.042) No. obs 130,690 175,010 890,215 357,153 53,206 73,419 39,535 146,906 No. patients 27,157 37,110 186,415 74,473 11,292 15,403 8,316 30,212 Dependent variable: ‘SwitchOut’. Only patients living in cities with more than 50,000 inhabitants were included. The seven left columns are from sample 1, the far right column is from sample 2. For ‘Others’, ‘Pat_BirthYear’ was replaced with a dummy variable equal to 0 for patients born in 1940 and equal to 1 for patients born in 1970. Each patient was observed up to five times. Sigma_u denotes the estimated standard deviation of the random patient-specific constant terms. Stata 13, the xtlogit procedure, was used in the estimations. Estimates with two-sided p-values < 1% are in boldface Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 8 of 10 instance, a high share of DT1 patients increased the patients with the same diagnosis; for instance, ‘Arth_share’ switch-out for patients with arthritis (meaning, patients has a negative effect (−15.032) for patients with arthritis, with arthritis were more likely to switch-out if their GPs and ‘Asthm_share’ has a negative effect (−10.406) for had more patients with DT1). The cross share effect was patients with asthma. Again, this may be the result of GPs not generally symmetric as a high share of patients with informally specializing in certain types of patients with arthritis reduced the switch-out for patients with DT1. chronic diseases. It may also result from the GPs’ general For all GP and patient characteristics, the significant qualities such as organizational skills, communication coefficients had the same sign across all patient groups. abilities, or empathic attitudes. It has been suggested that Patients tended to switch less often from GPs who had such patterns may result from patients’ negative interac- long patient lists (‘Ln_ListLength’) or who were specialists tions with healthcare providers, so that, for instance, obese in general medicine (‘GP_Specialist’). For older, female patients search for “obese friendly” physicians [25]. GPs, patients tended to switch out more often (‘GP_Age’). Patients could also make use of informal conversations This effect was even stronger for male GPs, for which the (word-of-mouth) with family, friends, or colleagues that full effect of age is obtained by adding the coefficients of recommend one GP or another, which seems to have a ‘GP_age’ and the interaction between a GP’sage andsex greater effect on the choice of GP than public information (‘GP_Age*GP_Sex’). disclosure [20]. The relationship between the GP and Patients born more recently (i.e., lower ‘Pat_BirthYear’) patient could also be a factor in patient choice, since or who had more comorbidities (‘Pat_Comorb’) tended chronic patients spend more time in primary care and to switch GPs more often. The 1% of patients who most would change their GP if they were not satisfied [3, 4]. We frequently used primary care (i.e., ‘Pat_Visits_dum’ =1) can assume that GPs who have high numbers of patients tended to switch less often than patients who had fewer with a particular disease might have a particular practice visits. However, among the remaining 99% of patients, style, which also attracts these patients, but these mecha- those with a higher number of primary care visits nisms may be complex, for instance for patients with (‘Pat_visits_win’)tendedtoswitchmoreoften. schizophrenia. In Table 5, the only exception from the The patient-specific effects are assumed to be normally general pattern is for patients with schizophrenia, for distributed, with a zero mean and an estimated standard which the effect of ‘Schi_share’ is insignificant. However, deviation, Sigma_u. For patients with arthritis, the value all other patient groups tend to disenroll more from GPs of Sigma_u can be interpreted as the difference in log- with high shares of patients with schizophrenia, poten- odds between a patient who has a patient-specific inter- tially suggesting that these GPs are less popular in general, cept one standard deviation from the mean (0.718) and and this may perhaps counter the “own share effect” a patient with an intercept equal to the mean value among patients with schizophrenia. (zero). This is about four times the numerical value of We find that all or most patient groups tend to disen- the coefficient for patient sex, and it corresponds to a roll less from GPs who have high shares of patients with difference in odds ratio equal to 2.050. In all patient arthritis, depression, and asthma. We assume that this groups, the estimated value for Sigma_u indicates that disenrollment pattern happens due to qualities of GPs the unobserved patient characteristics have a comparably that attract most patients, such as good communication large influence on disenrollment. and care coordination skills. For chronic patients who are intensive users of primary care it is important to find Discussion a GP that fits their needs, so they might change until Our data indicate that patients with chronic diseases are they find the right match. Patients in the comparison not allocated to GPs by chance alone (Fig. 2). One ex- group have, per se, no obvious reason to prefer GPs who planation could be that some GPs informally specialize, specialize in any chronic disease, but it is likely they have for example in DT2, and thus are able to establish and preferences regarding GP qualities. Thus, our finding maintain a “stock” of such patients. In so doing, the that in some cases the preferences of the comparison patient comorbidity shown in Table 3 would imply a group and of the patients with chronic diseases align tendency for these GPs to also have relatively higher suggests that GPs’ shares of chronic patients reveals shares of patients with arthritis and asthma. Moreover, information about these GPs’ general qualities. patients with chronic diseases tend to have comorbidities, A puzzling finding is that all or most patient groups contributing to their GPs having shares of patients with tend to disenroll more from GPs who have high shares of different diagnoses. This could partly explain why the patients with DT1 and schizophrenia. According to proportions of chronic disease types are all positively Norwegian guidelines, these two patient groups’ follow-up correlated, as shown in Table 4. happens in secondary care, in contrast to our other patient The coefficients in Table 5 suggest that chronic patients groups. Patients who receive follow-up in secondary care disenroll less often from GPs who have a high share of could perhaps be more indifferent to which GP they visit Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 9 of 10 for other acute illnesses. If so, they may be satisfied with (‘DT1_share’ and ‘DT2_share’), and the results for sub- GPs who have a practice style favoring patients who can samples defined for patients with DT1 and DT2. Second, be treated expediently over patients who need long-term our data did not include potentially relevant patient follow-up. With this interpretation, the high disenrollment variables such as cultural background, native language, among patients with schizophrenia (Table 2) can be income, educational background, or marital status. Disease interpreted not necessarily as a search for a GP who is severity and proper control of symptoms could also in- well-suitedforhandlingissues relatedtoschizophrenia fluence disenrollment behavior. To an extent, our ran- but perhaps as an expression of other, shorter-term dom effect logistic regressions can account for time- considerations. invariant patient variables, but future studies should GP specialization in general medicine has a negative consider including more variables in order to assess relationship with disenrollment, suggesting that patients their influence. Additional information about the GPs, prefer to stay with specialized GPs. List length also has a such as cultural background, length of time in practice, negative relationship with disenrollment for all patient and professional interests would also have been of groups, except for patients with schizophrenia. Previous interest. Third, the age distribution differs between our studies have found that non-chronic patients stay with selected comparison group, sample 2, and our main GPs with shorter patient lists, meaning that they value sample of interest, sample 1. Sample 2’sage distribution accessibility [10–12], in contrast to chronic patients who also differs from the age distribution across all groups value long patient lists, which is associated with higher in the full population without our specified chronic disease detection [13]. GP’s age is positively related with diseases.Thismeansthatthe estimatesfor sample 2in disenrollment for all patient groups, suggesting that pa- Tables 2 and 5 are likely to be biased, if interpreted as tients in general may prefer younger GPs. This effect of estimates for the full population. We believe that the age is supported by earlier findings [12]. For patients qualitative aspects of these results would not be very with arthritis, asthma, depression or DT2, this tendency different in the full population, but this is of course a is stronger for male than female GPs, perhaps because conjecture. Future register-based studies should consider there are fewer women among older GPs than among obtaining a comparison group with similar age distribu- younger GPs. In most patient groups, disenrollment was tion as the sample of main interest, for instance by not significantly associated with GP sex, except patients drawing patients randomly from the entire population. with asthma and depression, who tend to less often The data sets used in our logistic regressions were re- disenroll from male GPs. stricted with respect to municipality size. In smaller In all groups of patients with chronic disease, disen- municipalities, patient options for disenrollment will be rollment increased with the number of comorbidities. more limited by the fact that there are fewer local GPs This is consistent with the discussion above, given that to choose from. It is likely that including patients irre- management of patients with comorbidities is challen- spective of municipality size would yield estimated ging for primary care providers [27]. Our selection of effects less pronounced than those reported here – that patient groups was not, however, designed to investigate is, compared to the full population, our result are likely the effect of comorbidities in particular. Future studies to be biased away from zero. We also excluded observa- should consider including other diagnoses, such as cardio- tions where observed disenrollment seemed to be due vascular disease and cancer. A higher number of visits to to causes other than patients’ preferences for GPs. primary care also tended to increase disenrollment, but Patients and GPs who move, or GPs who retire or die, the negative coefficients for the dummy variable, identi- are likely to have demographic characteristics (e.g., age) fying patients who had more than 23 visits in a six that differ systematically from the distributions in the month period, may indicate that the relationship full patient and GP populations. It is more difficult to between disenrollment and the number of visits is not predict how including these observations would have linear. Younger patients generally disenroll more often influenced our results, but it would at least have com- and, except for patients with epilepsy and other patients plicated the interpretations. (sample 2), male patients disenroll less often. This study has three main imitations: first, although Conclusions the majority of the numerical data seemed reliable, we The following conclusions can be drawn from our found that as many as 77.8% of patients with DT1 were findings: 1) patients with chronic diseases are not allo- also registered as having DT2. Such “double diabetes” cated to GPs only by chance; 2) chronic patients that cases are not uncommon [28, 29], but it is likely that use primary care intensively disenroll less often from most of the cases in our data are due to diagnostic un- GPs who have a high share of patients with the same certainty or registration errors. This may affect both the diagnosis; and 3) most patient groups tend to remain results related to the share of patients with diabetes with GPs with a greater share of arthritis, asthma, and Mokienko and Wangen BMC Family Practice (2016) 17:170 Page 10 of 10 depression patients, which can indicate better quality 5. Saultz JW, Lochner J. Interpersonal continuity of care and care outcomes: a critical review. Ann Fam Med. 2005;3(2):159–66. care for these and other patient groups. These conclu- 6. Safran DG, et al. Switching doctors: predictors of voluntary disenrollment sions are distinct from the findings in the literature. from a primary physician's practice. J Fam Pract. 2001;50(2):130–6. To investigate this further, more objective quality mea- 7. Nagraj S, et al. Changing practice as a quality indicator for primary care: analysis of data on voluntary disenrollment from the English GP Patient surements should be obtained, such as adherence to Survey. BMC Fam Pract. 2013;14:89. treatment guidelines, surveillance of treatment outcomes 8. Kasteler J, et al. Issues underlying prevalence of “doctor-shopping” behavior. for chronic patients, and user satisfaction in general. If J Health Soc Behav. 1976;17(4):329–39. 9. Gray PG, Cartwright A. Choosing and changing doctors. Lancet. 1953; objective quality differences are found, further assess- 265(6799):1308–9. ments could be warranted, for instance, whether the 10. Marquis MS, Davies AR, Ware Jr JE. Patient satisfaction and change in current reimbursement system has an appropriate medical care provider: a longitudinal study. Med Care. 1983;21(8):821–9. 11. Rubin G, et al. Preferences for access to the GP: a discrete choice balance between capitation and fee-for service, or experiment. Br J Gen Pract. 2006;56(531):743–8. whether capitation should be risk-adjusted based on 12. Campbell JL, et al. Accessing primary care: a simulated patient study. Br J shares of patient types. Gen Pract. 2013;63(608):e71–6. 13. Campbell SM, et al. Identifying predictors of high quality care in English general practice: observational study. BMJ. 2001;323(7316):784–7. Abbreviations 14. Heje HN, et al. Doctor and practice characteristics associated with DT1: Type 1 diabetes; DT2: Type 2 diabetes; GP: General practitioner; differences in patient evaluations of general practice. BMC Health Serv Res. KUHR: Control and payment of reimbursements to health service providers 2007;7:46. (Kontroll og Utbetaling av HelseRefusjon) 15. Anwar MS, et al. Chronic disease detection and access: does access improve detection, or does detection make access more difficult? Br J Gen Pract. Acknowledgements 2012;62(598):e337–43. We wish to thank Tor Iversen for many valuable comments. The Norwegian 16. Soljak M, et al. Variations in cardiovascular disease under-diagnosis in Directorate of Health provided data. England: national cross-sectional spatial analysis. BMC Cardiovasc Disord. 2011;11:12. Funding 17. Norbury M, et al. Time to care: tackling health inequalities through primary The Research Council of Norway, project no. 204677/H10, provided funding. care. Fam Pract. 2011;28(1):1–3. 18. Mercer SW, Watt GC. The inverse care law: clinical primary care encounters Availability of data and materials in deprived and affluent areas of Scotland. Ann Fam Med. 2007;5(6):503–10. The datasets analyzed during the current study were obtained from the 19. Santos R, Gravelle H, and Propper C. Does quality affect patients’ choice of Norwegian Directorate of Health. Due to legal restrictions we are prevented doctor? Evidence from England. Econ J. 2016. doi:10.1111/ecoj.12282. from making the data publicly available or otherwise sharing individual level 20. Marshall MN, Shekelle PG, Leatherman S, Brook RH. The public release of data. For access to such data for research purposes, please contact the performance data: what do we expect to gain? A review of the evidence. Norwegian Directorate of Health directly. Jama. 2000;283(14):1866–74. 21. Marshall M, McLoughlin V. How do patients use information on health Authors’ contributions providers? BMJ. 2010;341:c5272. AM conceived the study, prepared the data and performed the statistical 22. Lurås H. The association between patient shortage and patient satisfaction calculations. Both AM and KRW participated in the study design, with general practitioners. Scand J Prim Health Care. 2007;25(3):133–9. interpretation of results, and writing of the manuscript. The final manuscript 23. Williams SJ, Calnan M. Key determinants of consumer satisfaction with has been read and approved by both AM and KRW. general practice. Fam Pract. 1991;8(3):237–42. 24. Sitzia J, Wood N. Patient satisfaction: a review of issues and concepts. Soc Competing interests Sci Med. 1997;45(12):1829–43. There author declare that there have no competing interests. 25. Gudzune KA, Bleich SN, Richards TM, Weiner JP, Hodges K, Clark JM. Doctor shopping by overweight and obese patients is associated with increased Consent for publication healthcare utilization. Obesity. 2013;21(7):1328–34. The patient data were obtained from public registers and patient consent for 26. Kristjansson E, et al. Predictors of relational continuity in primary care: publication was not required. patient, provider and practice factors. BMC Fam Pract. 2013;14:72. 27. Wallace E, et al. Managing patients with multimorbidity in primary care. Ethics approval and consent to participate BMJ. 2015;350:h176. The patient data were obtained from public registers and patient consent to 28. Pozzilli P, et al. Obesity, autoimmunity, and double diabetes in youth. participate was not required. The project has been approved by the Regional Diabetes Care. 2011;34 Suppl 2:S166–70. Committees for Medical and Health Research Ethics (REC South East), project 29. Cleland SJ, et al. Insulin resistance in type 1 diabetes: what is ‘double no. 2011/1708. diabetes’ and what are the risks? Diabetologia. 2013;56(7):1462–70. Received: 21 June 2016 Accepted: 2 December 2016 References 1. Starfield B. Commentary on regular primary care lowers hospitalisation risk and mortality in seniors with chronic respiratory disease. J Gen Intern Med. 2010;25(8):758–9. 2. Harris MF, Zwar NA. Care of patients with chronic disease: the challenge for general practice. Med J Aust. 2007;187(2):104–7. 3. Starfield B, Lemke KW, Herbert R, Pavlovich WD, Anderson G. Comorbidity and the use of primary care and specialist care in the elderly. Ann Fam Med. 2005;3(3):215–22. 4. Kaplan SH, Greenfield S, Ware Jr JE. Assessing the effects of physician- patient interactions on the outcomes of chronic disease. Med Care. 1989; 27(3 Suppl):S110–27.

Journal

BMC Family PracticeSpringer Journals

Published: Dec 15, 2016

There are no references for this article.