Per Hjerpe

A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena

Study objective: In social epidemiology, it is easy to compute and interpret measures of variation in multilevel linear regression, but technical difficulties exist in the case of logistic regression. The aim of this study was to present measures of variation appropriate for the logistic case in a didactic rather than a mathematical way. Design and participants: Data were used from the health survey conducted in 2000 in the county of Scania, Sweden, that comprised 10 723 persons aged 18–80 years living in 60 areas. Conducting multilevel logistic regression different techniques were applied to investigate whether the individual propensity to consult private physicians was statistically dependent on the area of residence (that is, intraclass correlation (ICC), median odds ratio (MOR)), the 80% interval odds ratio (IOR-80), and the sorting out index). Results: The MOR provided more interpretable information than the ICC on the relevance of the residential area for understanding the individual propensity of consulting private physicians. The MOR showed that the unexplained heterogeneity between areas was of greater relevance than the individual variables considered in the analysis (age, sex, and education) for understanding the individual propensity of visiting private physicians. Residing in a high education area increased the probability of visiting a private physician. However, the IOR showed that the unexplained variability between areas did not allow to clearly distinguishing low from high propensity areas with the area educational level. The sorting out index was equal to 82%. Conclusion: Measures of variation in logistic regression should be promoted in social epidemiological and public health research as efficient means of quantifying the importance of the context of residence for understanding disparities in health and health related behaviour.

Validity of registration of ICD codes and prescriptions in a research database in Swedish primary care: a cross-sectional study in Skaraborg primary care database

BackgroundIn recent years, several primary care databases recording information from computerized medical records have been established and used for quality assessment of medical care and research. However, to be useful for research purposes, the data generated routinely from every day practice require registration of high quality. In this study we aimed to investigate (i) the frequency and validity of ICD code and drug prescription registration in the new Skaraborg primary care database (SPCD) and (ii) to investigate the sources of variation in this registration.MethodsSPCD contains anonymous electronic medical records (ProfDoc III) automatically retrieved from all 24 public health care centres (HCC) in Skaraborg, Sweden. The frequencies of ICD code registration for the selected diagnoses diabetes mellitus, hypertension and chronic cardiovascular disease and the relevant drug prescriptions in the time period between May 2002 and October 2003 were analysed. The validity of data registration in the SPCD was assessed in a random sample of 50 medical records from each HCC (n = 1200 records) using the medical record text as gold standard. The variance of ICD code registration was studied with multi-level logistic regression analysis and expressed as median odds ratio (MOR).ResultsFor diabetes mellitus and hypertension ICD codes were registered in 80-90% of cases, while for congestive heart failure and ischemic heart disease ICD codes were registered more seldom (60-70%). Drug prescription registration was overall high (88%). A correlation between the frequency of ICD coded visits and the sensitivity of the ICD code registration was found for hypertension and congestive heart failure but not for diabetes or ischemic heart disease.The frequency of ICD code registration varied from 42 to 90% between HCCs, and the greatest variation was found at the physician level (MORPHYSICIAN = 4.2 and MORHCC = 2.3).ConclusionsSince the frequency of ICD code registration varies between different diagnoses, each diagnosis must be separately validated. Improved frequency and quality of ICD code registration might be achieved by interventions directed towards the physicians where the greatest amount of variation was found.

Gender differences in antihypertensive drug treatment: results from the Swedish Primary Care Cardiovascular Database (SPCCD)

There are gender differences in antihypertensive treatment. This study aimed to investigate if gender differences in treatment could be explained by comorbidities. In addition, we aimed to study whether blood pressure control is different in women and men, and whether women interrupt treatment more often with angiotensin-converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs) compared with men.This cohort study within the Swedish Primary Care Cardiovascular Database included 40,825 patients with hypertension attending primary health care from 2007 to 2008. Cardiovascular comorbidities, with the exception of heart failure, were more common in men. Women were more often treated with diuretics, and men with ACEI, as were hypertensive patients with diabetes. Comorbidities could not entirely explain gender differences in antihypertensive treatment in a regression model. Women had higher systolic and lower diastolic blood pressure; this was also true in subgroups with cardiovascular comorbidity. Men more often than women were prescribed ACEIs/ARBs and interrupted treatment. Women and men are treated with different antihypertensive drugs, and this is not fully explained by differences in comorbidities. Women have higher systolic blood pressures, irrespective of comorbidity. Men have interrupted treatment more often with ACEIs/ARBs. These gender differences could affect outcome and warrant further investigation.

The Swedish Primary Care Cardiovascular Database (SPCCD): 74 751 hypertensive primary care patients.

Abstract Objective. To describe the Swedish Primary Care Cardiovascular Database, SPCCD. Design. Longitudinal data from electronic medical records, linked to national registers. Setting. 48 primary healthcare centres in urban (south-western Stockholm) and rural (Skaraborg) regions in Sweden. Subjects. Patients diagnosed with hypertension 2001–2008. Main outcome measures. Blood pressure (BP) and impact of retrieval of data on BP levels, clinical characteristics, co-morbidity and pharmacological treatment. Results. The SPCCD contains 74 751 individuals, 56% women. Completeness of data ranged from > 99% for drug prescriptions to 34% for smoking habits. BP was recorded in 98% of patients during 2001–2008 and in 63% in 2008. Mean BP based on the last recorded value in 2008 was 142 ± 17/80 ± 13 mmHg. Digit preference in BP measurements differed between the two regions, p < 0.001. Antihypertensive drugs were prescribed in primary healthcare to 88% of the patients in 2008; however, when all prescribers were included 96% purchased their drugs. Cardiovascular co-morbidity and diabetes mellitus were present in 28% and 22%, respectively. Conclusion. This large and representative database shows that there is room for improvement of BP control in Sweden. The SPCCD will provide a rich source for further research of hypertension and its complications.

Diagnosis of patients with raised serum calcium level in primary care, Sweden

Objective. To study the diagnosis of hypercalcaemic patients and to evaluate whether frequent analyses of serum calcium can detect more patients with hypercalcaemia. Design. Retrospective study of serum calcium analyses performed during the time period 1992–2000 and of the medical records of patients with elevated serum calcium levels between 1995 and 2000. Setting. Primary care in Tibro, Sweden. Subjects. Patients from the local community attending the primary healthcare centre. Main outcome measures. Frequency of serum calcium analyses, hypercalcaemic patients, and their diagnosis. Results. Doubling the number of serum calcium analyses did not increase the detected number of raised calcium levels. On the other hand, more frequent parathyroid hormone (PTH) analyses resulted in a corresponding increase in detected high PTH levels. In Tibro, 15% (n = 22) of the patients with hypercalcaemia were diagnosed with primary hyperparathyroidism, giving a rate of 0.22%. This is comparable to the prevalence in other population studies. Over 40% (n = 9) of patients with primary hyperparathyroidism in the study had only slightly raised serum calcium levels (2.55–2.60 mmol/l). In 70% (n = 99) of the cases, the cause of hypercalcaemia was unknown. The second most common diagnosis was skeletal disorders followed by kidney disease. Conclusion. An increase in the number of serum calcium analyses did not result in increased detection of raised calcium levels. In contrast, an increase in the number of PTH analyses resulted in increased detection of primary hyperparathyroidism. Therefore, PTH analyses should be used more frequently.

No physician gender difference in prescription of sick-leave certification: A retrospective study of the Skaraborg Primary Care Database

Abstract Objective. The primary objective was to investigate how physicians’ gender and level of experience affects the rate and length of sick-leave certificate prescription. The secondary objective was to study the physicians’ gender and professional experience in relation to the diagnoses on the certificates. Design. Retrospective, cross-sectional study of computerized medical records from 24 health care centres in 2005. Setting. Primary care in Sweden. Subjects. Primary care physicians (n = 589) and patients (n = 88 780) aged 18–64 years. Main outcome measures. Rate and duration of sick leave certified by different categories of physicians and for different diagnoses and gender of patients. Results. Sick leave was certified in 9.0% (musculoskeletal (3%) and psychiatric (2.3%) diagnoses were most common) of all contacts and the mean duration was 32.2 days. Overall there was no difference between male and female physicians in the sick-leave certification prescription rate (9.1% vs. 9.0%) or duration of sick leave (32.1 vs. 32.6 days). The duration of sick leave was associated with the physicians level of professional experience in general practice (GPs (Distriktläkare) 37, GP trainees (ST-läkare) 26, interns (AT-läkare) 20 and locum (vikarier) 19 days, p < 0.001). Conclusion. Contrary to earlier studies we found no difference in sick-leave certification prescription rate and length between male and female physicians.

Understanding adherence to therapeutic guidelines: a multilevel analysis of statin prescription in the Skaraborg Primary Care Database

PurposeIn Skaraborg, Sweden, the economic responsibility for tax-financed prescription drug costs was transferred from the regional administrative level to the local level (health care centre; HCC) in 2003. The aim of this study was to investigate the impact of this decentralization of economic responsibility on adherence to guidelines for prescribing lipid-lowering drugs.MethodsData from all 24 public HCCs in Skaraborg on prescriptions for lipid-lowering drugs during 2003 and 2005 were extracted from the Skaraborg Primary Care Database (SPCD). Multilevel regression analysis (MLRA) was used to disentangle the variances at different levels of data (patient, physician, HCC). The outcome variable on the patient level was the prescription of the recommended statin (yes/no). Sex and age of the patients and sex, age and occupational status of the physician were included as fixed effects. The variance was expressed as the median odds ratio (MOR).ResultsThe prevalence of adherence to guidelines for the prescription of statins increased from 77% in 2003 to 84% in 2005. The MLRA showed that in 2003 the variance was equally distributed between the HCC and physician levels (MORHCC2003 = 1.89 vs. MORPHYSICIAN2003 = 1.88). The variance between physicians and between HCCs decreased considerably between 2003 and 2005. The inclusion of individual and physician characteristics did not explain any of the remaining variance.ConclusionThe decentralized budget appears to have increased adherence to guidelines and reduced inefficient variation in prescribing.

Antihypertensive control and new-onset atrial fibrillation: Results from the Swedish Primary Care Cardiovascular Database (SPCCD):

Background Atrial fibrillation is associated with increased cardiovascular morbidity and mortality. Hypertension is an important risk factor for the development of atrial fibrillation. Aim This study assessed the relationship between blood pressure control and new-onset atrial fibrillation in hypertensive patients. Methods and results We followed 45,530 hypertensive patients with no previously documented atrial fibrillation, attending primary healthcare in Sweden during 2001–2008. After a mean follow-up of 3.5 years 2057 patients (4.5%) developed atrial fibrillation. Compared to patients with no atrial fibrillation, the new-onset atrial fibrillation group (after adjustment for age, sex, diabetes mellitus, heart failure, ischaemic heart disease, cerebrovascular disease and number of visits) had higher mean in-treatment systolic blood pressure (SBP) and diastolic blood pressure of 3.8 mmHg (95% confidence interval (CI) 3.0–4.6; P < 0.0001) and 1.6 mmHg (95% CI 1.2–2.0; P < 0.0001), respectively. Similarly, mean in-treatment pulse pressure in the new-onset atrial fibrillation group was 2.2 mmHg (95% CI 1.6–2.9; P < 0.001) higher. In a logistic regression analysis, achieved SBP ≥ 140 mmHg was associated with a higher risk of new-onset atrial fibrillation, compared to SBP 130–139 mmHg (odds ratio (OR) 1.5; 95% CI 1.3–1.7) and to SBP < 130 mmHg (OR 1.3; 95% CI 1.1–1.5). There was no difference in risk for new-onset AF between SBP 130–139 and <130 mmHg (OR 0.9; 95% CI 0.7–1.1). Conclusion The present findings indicate that blood pressure control in hypertension is associated with a lower risk of new-onset atrial fibrillation.

Long-term follow-up of patients with elevated serum calcium concentrations in Swedish primary care

Abstract Objective. To follow up patients with elevated calcium concentrations after 10 years. Design. Longitudinal, using medical records, questionnaires, and clinical investigation. Setting. Primary care in Tibro, Sweden, 2008–2010. Subjects. 127 patents with elevated calcium concentrations and 254 patients with normal calcium concentrations from the local community, attending the health care centre. Main outcome measures. Diagnoses and mortality in patients with elevated calcium concentrations in 1995–2000, compared with patients with normal calcium concentrations and the background population. Results. The proportion of patients for whom no underlying cause was detected decreased from 55% at baseline to 12% at follow-up. Primary hyperparathyroidism was most common in women, 23% at baseline and 36% at follow-up, and the cancer prevalence increased from 5% to 12% in patients with elevated calcium concentration. Mortality tended to be higher in men with elevated calcium concentrations compared with men with normal calcium concentrations, and was significantly higher than in the background population (SMR 2.3, 95% CI 1.3–3.8). Cancer mortality was significantly increased in men (p = 0.039). Low calcium concentrations were also associated with higher mortality (p = 0.004), compared with patients with normal calcium concentrations. Conclusion. This study underscores the importance of investigating patients with increased calcium concentrations suggesting that most of these patients – 88% in our study – will turn out to have an underlying disease associated with hypercalcaemia during a 10-year follow-up period. Elevated calcium concentrations had a different disease pattern in men and women, with men showing increased cancer mortality in this study.

Increased registration of hypertension and cancer diagnoses after the introduction of a new reimbursement system

Abstract Objective. To investigate the impact on ICD coding behaviour of a new case-mix reimbursement system based on coded patient diagnoses. The main hypothesis was that after the introduction of the new system the coding of chronic diseases like hypertension and cancer would increase and the variance in propensity for coding would decrease on both physician and health care centre (HCC) levels. Design. Cross-sectional multilevel logistic regression analyses were performed in periods covering the time before and after the introduction of the new reimbursement system. Setting. Skaraborg primary care, Sweden. Subjects. All patients (n = 76 546 to 79 826) 50 years of age and older visiting 468 to 627 physicians at the 22 public HCCs in five consecutive time periods of one year each. Main outcome measures. Registered codes for hypertension and cancer diseases in Skaraborg primary care database (SPCD). Results. After the introduction of the new reimbursement system the adjusted prevalence of hypertension and cancer in SPCD increased from 17.4% to 32.2% and from 0.79% to 2.32%, respectively, probably partly due to an increased diagnosis coding of indirect patient contacts. The total variance in the propensity for coding declined simultaneously at the physician level for both diagnosis groups. Conclusions. Changes in the healthcare reimbursement system may directly influence the contents of a research database that retrieves data from clinical practice. This should be taken into account when using such a database for research purposes, and the data should be validated for each diagnosis.