OUP user menu

Quality of stroke prevention in general practice: relationship with practice organization

Johan S. de Koning, Niek Klazinga, Peter J. Koudstaal, AD Prins, Gerard J. J. M. Borsboom, Johan P. Mackenbach
DOI: http://dx.doi.org/10.1093/intqhc/mzi004 59-65 First published online: 24 January 2005

Abstract

Objective. To investigate the relationship between elements of practice organization related to stroke prevention in general practice, and suboptimal preventive care preceding the occurrence of stroke.

Design. This study was conducted among 69 Dutch general practitioners in the Rotterdam region. Information on the implementation of elements of practice organization related to stroke prevention was collected by postal questionnaire. Data on the process of patient care were collected by means of chart review and interviews with general practitioners. Cases of stroke (n = 186) were retrospectively audited by an expert panel with guideline-based review criteria. Using logistic regression analysis we investigated the relationship between the probability of suboptimal care delivery and the presence of specific elements of practice organization related to stroke prevention (tailored information systems, formal delegation of preventive tasks, standardization of care).

Results. For some elements of practice organization significant relationships with the quality of stroke prevention were found. Suboptimal care was less common among general practitioners with a higher level of noting high risk patients in the patient records (odds ratio 0.30; 95% CI 0.13–0.69, P = 0.01), delegating follow-up visits to support staff (odds ratio 0.42; 95% CI 0.22–0.82, P = 0.01) and compliance with the hypertension guideline (odds ratio 0.57; 95% CI 0.41–0.78, P = <0.001). Except for practice type (general practitioners in health centres less often provided suboptimal care, P = 0.02), no significant relationships with general practitioner and practice characteristics were found.

Conclusion. This study shows that general practitioners with a higher level of integrated organizational structures for stroke prevention (record keeping, formal delegation of preventive tasks, guideline compliance) are less likely to deliver suboptimal care.

  • audit
  • general practice
  • quality assessment
  • stroke prevention

Stroke continues to be the third leading cause of death, after heart disease and cancer, and is a major cause of long-term disability in industrialized countries [1]. Given its significant impact on public health and the fact that there are no effective treatments for most types of stroke, preventive strategies are of utmost importance and offer the greatest potential for reducing the burden of this disease [2].

In many health care systems, general practice plays a prominent role in prevention. Of all health care providers, general practitioners have the most frequent contact with their patients and, therefore, have easy access to individuals at risk of cardiovascular disease. It is for this reason that in primary care, general practitioners receive increasing support to improve the quality of cardiovascular disease prevention.

Since the early 1990s, a more systematic approach towards prevention has been propagated by the Dutch Ministry of Health, Welfare, and Sports and professional organizations for general practitioners [3]. In addition to advocating a more proactive approach towards cardiovascular disease prevention, modernization of practice organization with regard to patient information systems, task delegation, and standardization of care was recommended. Since then, general practitioners have invested in the development and implementation of practice structures to systematically enhance the quality of cardiovascular disease prevention. These strategies and activities have become part and parcel of the overall quality improvement policies and programmes for systematic quality assurance and improvement in general practice in the Netherlands (e.g. national guideline programme, obligatory peer review, continuing medical education courses) [4].

It has not yet, however, been demonstrated whether structural adaptations in practice organization measurably improve the general practitioner’s quality of preventive care delivery compared with colleagues in practices with less organized systems for prevention. In this study, we investigated the relationship between elements of practice organization considered to be relevant to stroke prevention in general practice, and suboptimal preventive care preceding the occurrence of stroke.

Methods

Design

The study was conducted within the framework of an audit study on quality of stroke prevention in general practice in the city of Rotterdam and surrounding region (the Netherlands) [5]. After approval from the local medical ethics committee, patients were identified and selected from the two principle referral hospitals for all strokes that occur in the region. The study was restricted to patients aged 39–80 years with a first-ever stroke. Within this age category the occurrence of stroke is assumed to reflect potential deficiencies in quality of care [6]. Patient records were searched to identify patients’ general practitioner. Of the 77 general practitioners participating in the audit study, 69 general practitioners were included in the present study (69 general practitioners returned the postal questionnaire with questions addressing the implementation of elements of practice organization).

Data collection

Data on the process of care at patient level were collected by means of structured face-to-face interviews with the general practitioner, using separate questionnaires for each stroke patient. At the time of the interview the general practitioner used either hand-written or electronic patient records to retrieve the patient’s information. The questionnaire comprised questions related to patient characteristics, medical and family history of cardiovascular risk factors, and the detection and treatment of risk factors for stroke (e.g. hypertension, diabetes mellitus, transient ischemic attack, and cardiac failure). Similarly, information on lifestyle-related risk factors (e.g. cigarette smoking, overweight, and alcohol consumption) was collected.

In addition to the interview, general practitioners received a postal questionnaire containing 77 pre-structured questions on elements of practice organization (systematic recording of preventive activities, risk factors and risk groups, support staff and delegation of preventive activities, knowledge and application of clinical practice guidelines, continuing medical education, formalized co-operation within general practitioner practice, and formalized co-operation with other health care providers). Questions addressing general practitioners’ personal characteristics (age, sex, year of qualification, teaching practice, working hours) and other practice characteristics (practice type, practice location, list size) were also included.

Assessment of quality of care

The quality of preventive care was based on the judgment of a six-member panel of experts, comprising three neurologists and three general practitioners. The panel used six clinical practice guidelines relevant to stroke prevention (hypertension, diabetes mellitus, transient ischemic attack, peripheral vascular disease, cardiac failure, and angina pectoris) to assess the quality of care. These guidelines were developed and implemented by the Dutch College of General Practitioners as part of a national guideline programme [7]. From each guideline, the panelists identified specific elements of care and systematically converted them into review criteria (n = 65) [8].

Panelists were given detailed information on the process of care delivery. Based on the identified aspects of suboptimal care and seriousness of shortcomings (minor versus major), panelists allocated grades on a scale of 0 to 3 (Table 1). For example, if a general practitioner did not conform to recommendations for quarterly follow-up of treated hypertensive patients, hypertensive care was regarded as being suboptimal. This in itself justifies a grading of 1. If the patient’s blood pressure remained above the recommended target level (diastolic blood pressure >95 mmHg), this suboptimal care was considered to have failed to prevent stroke. Assignment of a grade 2 (‘possibly have failed’) or 3 (‘likely to have failed’) depended on the actual blood pressure level and/or general practitioner’s clinical performance. Only care under the general practitioner’s responsibility was included in the assessment.

View this table:
Table 1

Final grades given by the expert panel

GradingCases (n = 186)%
0No suboptimal factors have been identified10255
1Suboptimal factor(s) have been identified, but are unlikely to have failed to prevent the stroke of this patient2614
2Suboptimal factor(s) have been identified, and possibly have failed to prevent the stroke of this patient3820
3Suboptimal factor(s) have been identified, and are likely to have failed to prevent the stroke of this patient1810
No consensus reached21

The cases were assessed in a two-round evaluation, with a final plenary round. The panelists operated in subpanels comprising two persons each (each subpanel assessed approximately 60 patients). In the first round, the panelists individually evaluated the cases assigned to his/her subpanel. If, within a subpanel, equal grades were assigned to a particular case, no further evaluation was needed. If no consensus was reached, a second evaluation was done. During this round, the panelists received a copy of their own grading form and a copy of that of the other panelist. If no consensus decision was reached in the second evaluation round, the case was discussed in the final and plenary round.

For 36 cases, the inter-subpanel agreement was measured by Cohen’s κ statistic, which accounts for chance agreement [9]. These patients were selected by drawing a random sample of 12 cases out of all patients assessed by a particular subpanel. Of these cases, copies were made and sent for evaluation to another subpanel. The inter-subpanel agreement was κ = 0.63 (overall agreement on assigned grades between subpanels was 74%). In 55% of all cases, no suboptimal care (grade 0) was identified by the expert panel, whereas in 44%, suboptimal care was identified (grades 1–3) (Table 1). Deficiencies in hypertensive care and patients’ cardiovascular risk profile assessment were found most frequently.

Analysis

Logistic regression analysis was used to investigate the relationship between the probability of patients receiving optimal or suboptimal care and elements of practice organization for stroke prevention and characteristics of general practitioners. An important assumption of standard logistic regression analysis is the statistical independence of observations. Correlations between observations may result in estimated confidence intervals that are too narrow and consequently in too optimistic estimates of statistical significance. For many practices, data on several patients were included in the analysis. These patients have their general practitioner in common, resulting in correlations between observations, as care provided to patients from the same practice is likely to be more alike than care provided to patients from different practices. These correlations were taken into account by specifying a random intercepts ‘working correlation matrix’ in a Generalized Estimating Equations (GEE) logistic regression analysis available in proc genmod of SAS version 9.1. This approach allows for unbalanced data (i.e. unequal numbers of patients per general practitioner) so that data from all available patients (except the two for whom no consensus was reached on the final grading of quality of care) could be included in the analysis.

The rating scale for care delivery was dichotomized into a score of zero, indicating optimal care delivery, and a score of one, denoting the three grades (score 1, 2, or 3) of suboptimal care. For the analysis of adherence to practice guidelines, affirmative answers were counted. These counts were included as continuous linear variables in the logistic regressions relating adherence to quality of care.

We fitted simple models (one practice, one general practitioner characteristic, adherence to one guideline at a time) and multivariate models (all variables considered in a particular section of the analysis or those found statistically significant in the simple approach). However, to account for confounding influences of differing compositions of patient groups between practices, we adjusted all models for the most important patient characteristic, namely the presence of hypertension (in many cases suboptimal care was related to hypertension management). In separate logistic regression analyses we investigated whether patterns of missing values (0 if an observation was available, and 1 if an observation was missing) in the independent variables were related to quality of care.

Results

Study population

The 69 general practitioners included in the study provided data on 186 patients (average number of patients per general practitioner was 2.7). Twenty-four general practitioners had only one patient eligible for inclusion, 12 had two, 13 had three, 11 had four, and nine general practitioners had five patients or more. The participating general practitioners were comparable to other Dutch general practitioners for sex, age distribution, practice type (single and duo practices) and list size [10]. Of all general practitioners, 70% started their current practice in the period 1970–89 and 16% between 1990 and 1999.

Practice organization for stroke prevention

Overall, between 65% and 91% of the general practitioners reported implementation of at least one of the two conditions of delegation of preventive activities to the practice assistant (68%), participation in continuing medical education (65%), co-operation within practice teams (91%), and co-operation with other health care providers (71%). Sixty-nine per cent of the general practitioners had implemented three of the five conditions for systematic recording of risk factors or risk groups. With respect to systematic recording of risk factors and risk groups, the majority of general practitioners had noted diabetic patients, and had registered blood pressure measurements in their patient records. Systematic recording of patients with an elevated risk of cardiovascular disease and smoking status was done by less than half of the general practitioners (Table 2). However, 16% reported that they did not record smoking in their patients’ records, even if discussed with the general practitioner and found positive. Formal delegation of follow-up visits for hypertensive and diabetic patients to the practice assistant was done by about 50% of general practitioners. Of those who did not delegate follow-up visits of hypertensive patients on a regular basis, 28% never and 18% rarely delegated. A remarkably low percentage of general practitioners was involved in peer review activities.

View this table:
Table 2

Percentages of general practitioners reporting the implemention of specific elements of practice organization, and the relations between these elements of practice organization and suboptimal care delivery

Elements of practice organizationGPs (n = 69) %Patients (N = 186)
Simple model1Multivariate model2
OR (95% CI)P-valueOR (95% CI)P-value
Systematic recording in patient’s record
    Elevated risk of cardiovascular disease440.78 (0.34–1.79)0.56
    Diabetes mellitus830.34 (0.13–0.93)0.040.30 (0.13–0.69)0.01
    Blood pressure measurements760.22 (0.07–0.71)0.010.66 (0.23–1.85)0.43
    Smoking (if discussed with patient)411.89 (0.83–4.33)0.13
    Use of recording cards or electronic prevention module (HIS)240.62 (0.29–1.29)0.20
Formal delegation to practice assistant
    Follow-up visits of diabetic patients480.96 (0.48–1.91)0.91
    Follow-up visits of hypertensive patients540.44 (0.22–0.88)0.020.42 (0.22–0.82)0.01
Formalized co-operation with
    Dietitian700.39 (0.16–0.93)0.030.96 (0.50–1.84)0.20
    Diabetes nurse250.66 (0.34–1.28)0.22
Continuing medical education in
    Cardiology320.91 (0.41–2.01)0.81
    Diabetes mellitus560.53 (0.26–1.09)0.08
Co-operation within GP team
    Peer review422.61 (1.21–5.61)0.014.21 (1.49–11.9)0.01
    Formalized co-operation851.19 (0.40–3.54)0.75
  • Answers on the independent variables were always recorded as yes/no. The ‘no’ category was used as reference in all models.

  • 1 Odds ratios were adjusted for the patients’ diagnosis of hypertension. Number of observations per variable included in the models ranged from 169 (peer review) to 109 (systematic recording of diabetes mellitus and blood pressure).

  • 2 Odds ratios were adjusted for the patient diagnosis of hypertension and all variables statistically significant in the univariate analyses.

Relationship between elements of practice organization and quality of care

Systematic recording and formal task delegation.

Results of the separate logistic regression analyses and the multivariate model can be found in Table 2. Most of the odds ratios in the separate analyses were <1, indicating that positive answers generally implied decreased odds for delivering suboptimal care. However, only four of these relationships were statistically significant, namely systematic noting of diabetes mellitus and recording of blood pressure measurements in the patients’ record, delegation of follow-up visits of hypertensive patients to the practice assistant, and formalized co-operation with a dietitian. Results in the opposite direction were found for systematic recording of smoking and peer review (odds ratios indicating increased odds for delivering suboptimal care in the case of a positive answer). Inclusion in the multivariate model changed results in such a way that systematic recording of blood pressure measurements and formalized co-operation with a dietitian were no longer significant. The direction of the effect, however, remained unchanged. The logistic regressions in which we analysed the patterns of missing data did not yield evidence for a bias in the odds ratios as a result of selection effects.

Compliance with clinical practice guidelines.

Results of the logistic regression analyses relating guideline compliance to quality of care can be found in Table 3. Among the simple analyses, investigating one clinical practice guideline at a time, we found a significant relationship between compliance with the hypertension guideline and the quality of preventive care. The odds ratio indicated a substantial decrease in odds for every aspect of the guideline extra complied with by the general practitioner. The odds ratios for the remaining guidelines (except diabetes mellitus) pointed in opposite directions. These associations were statistically not significant. The overall findings in the simple analyses did not change when all variables were adjusted for each other in the multivariate model. As above, patterns of missing data did not bias these results.

View this table:
Table 3

Relationship between compliance with clinical practice guidelines and suboptimal care

Clinical practice guidelineNo. aspectsSimple model1Multivariate model2
Odds ratio (95% CI)P-valueOdds ratio (95% CI)P-value
Hypertension60.73 (0.54–0.98)0.040.57 (0.41–0.78)<0.001
Diabetes mellitus50.93 (0.61–1.40)0.720.77 (0.54–1.09)0.14
Angina pectoris31.32 (0.79–2.21)0.291.16 (0.66–2.01)0.61
Chronic heart failure31.20 (0.80–1.80)0.391.37 (0.85–2.20)0.20
Transient ischemic attack32.02 (0.64–6.41)0.232.06 (0.73–5.78)0.17
Peripheral vascular disease31.18 (0.37–3.77)0.792.17 (0.66–7.15)0.20
  • 1 Odds ratios were adjusted for the patient diagnosis of hypertension.

  • 2 Odds ratios were adjusted for the patient diagnosis of hypertension and all other variables in the table.

General practitioner and practice characteristics in relation to quality of care.

No evidence was found for the relationship between general practitioners’ characteristics (sex, year of qualification, working hours, and tutor) and the quality of care (Table 4). However, we found that general practitioners who qualified in the period 1975–84 had higher odds for delivering suboptimal care as compared with general practitioners qualified in the period 1960–74 or 1985–94 (borderline significant).

View this table:
Table 4

Relationship between general practitioner/practice characteristics and suboptimal care delivery

General practitioner/practice characteristicsSimple model1Multivariate model2
Odds ratio (95% CI)P-valueOdds ratio (95% CI)P-value
General practitioner
    Year of qualification0.050.07
        1960–740.87 (0.37–2.01)0.99 (0.44–2.24)
        1975–841.96 (0.92–4.20)2.27 (1.00–5.12)
        1985–94 (ref.)1.001.00
    GP tutor
        No (ref.)1.000.611.000.83
        Yes0.84 (0.42–1.68)0.92 (0.44–1.95)
    Working hours
        Part-time0.34 (0.10–1.15)0.080.76 (0.22–2.67)0.67
        Full-time (ref.)1.001.00
    Sex
        Male0.31 (0.06–1.57)0.160.32 (0.05–2.11)0.24
        Female (ref.)1.001.00
Practice
    Type of practice0.270.02
        Single-handed (ref.)1.001.00
        Duo practice3.33 (1.00–11.05)3.56 (0.99–12.78)
        Group practice1.08 (0.23–5.13)0.98 (0.26–3.73)
        Health centre1.03 (0.37–2.83)0.38 (0.15–0.94)
    Neighbourhood
        Deprived0.76 (0.38–1.52)0.440.69 (0.36–1.32)0.26
        Not deprived1.001.00
    List size (>2500 patients)
        <2500 patients1.50 (0.78–2.88)0.231.16 (0.58–2.34)0.67
        >2500 patients1.001.00
  • 1 Odds ratios were adjusted for the patient diagnosis of hypertension.

  • 2 Odds ratios were adjusted for all other variables in this column and for patient diagnosis of hypertension.

With respect to practice characteristics, general practitioners working in larger settings delivered suboptimal care less often (the high odds ratio of the category ‘duo practice’ in both simple and multivariate models, and the low and statistically significant odds ratio ‘health centre’ in the multivariate model were based on small numbers as indicated by the large confidence intervals. About 75% of the cases came from single-handed practices, and only 11 patients belonged to health centres). No significant relationships between practice location and list size and quality of care were observed.

Discussion

This study investigated the relationship between elements of practice organization, and the quality of stroke prevention in general practice. It shows that for some elements of practice organization a positive relationship exists with the quality of care. In particular, systematic noting of diabetes mellitus in patients’ records, formal delegation of preventive activities to the practice assistant, and guideline compliance were associated with the quality of general practitioners’ preventive behaviour.

Before further elaboration on the results, limitations of this study deserve to be mentioned. Deficiencies in care delivery identified by the expert panel might have over- or underestimated suboptimal care delivery. For example, suboptimal care might have been overestimated because (i) the panelists were aware of the severity of outcome and therefore performed a more critical analysis of care, and (ii) the presence of legitimate reasons for non-conformance with the guideline that were not included in the review criteria. This may imply that correct clinical behaviour was in fact erroneously classified as suboptimal care. On the other hand, suboptimal care may have been underestimated because of reporting of socially desirable behaviour by general practitioners when they did not adhere to a particular practice guideline. In balance, we expect that the number of stroke patients receiving suboptimal care failing to prevent the occurrence of stroke is most likely to be underestimated. The question remains whether a relationship exists between the degree of underestimation of suboptimal care and the presence of a practice organization relevant to stroke prevention. This would be the case if, for example, general practitioners who tend to overrate their actual professional performance, also overrate self-reported compliance to practice guidelines (leading to an overestimation of the relationship between compliance and suboptimal care delivery).

Regarding the relationship between recording of risk factors and risk groups and the quality of care delivery, our findings concur with those of other studies. Research findings indicate that for various cardiovascular disease risk factors such as hypertension (recording of blood pressure readings), cigarette smoking, and overweight, the level of recording in general practice does not reach optimal levels [11,,12]. Others have reported on the positive relationship between the organization of disease prevention in general practice and the recording behaviour of general practitioners. Practices with a better organized recording system appeared to have higher levels of risk factor recording than those without [13,,14]. In accordance with other studies, we report that improved practice of recording, particularly noting diabetes mellitus, in the patient’s record, results in better quality of care to patients with a high risk of developing the disease [15–17]. These results confirm that adequate information about a patient’s risk profile is essential in order to target preventive care effectively.

Systematic and regular delegation of preventive activities to the practice assistant is expected to improve quality of care as well [18]. General practitioners with a higher rate of delegation to the practice assistant spend more time per patient, which is considered an important determinant for quality of clinical performance [19]. These studies support our finding that general practitioners with a higher rate of delegation less often provide suboptimal care. Similar to other studies [20], our results show that educational intervention and peer review to improve the clinical behaviour of general practitioners have little positive impact on care delivery. Most studies report on the limited effectiveness of traditional continuing medical education in improving the delivery of preventive services in general practice. Apart from increasing the general practitioner’s awareness of a particular aspect of care, generally, traditional continuing medical education has not proved successful [21,,22]. In our study, peer review was significantly related to the quality of care, however, in the opposite direction. In a recent study on general practitioners’ clinical performance with respect to blood pressure control, Frijling et al. [23] concluded that general practitioner characteristics had little effect on clinical performance; our study confirms this result.

Although the analysis was based on relatively small numbers, we found significant variation in care between general practitioners working in different practice settings. General practitioners working in larger practices (health centres) less often provided suboptimal care as compared with their colleagues in single-handed practices. This finding supports policy initiatives in the Netherlands and the UK aimed at improving mechanisms in primary care that could protect quality standards. With respect to practice type and clinical performance, it is expected that general practitioners in single-handed practices, working in relative clinical isolation without ready support from colleagues, more often deliver suboptimal care. For this reason, general practitioners are encouraged to work in larger practices. Studies on differences in performance between group practices and single-handed practices, however, do not provide evidence that single-handed general practitioners are indeed under-performing [24]. There are indications that instead of practice type, practice location (deprived versus non-deprived areas) influences prevention in general practice [24]. This finding we confirm in our study on stroke prevention in primary care in deprived and non-deprived neighbourhoods [25].

Our findings support the assumption that policy initiatives to improve patient information systems, task delegation, and partnership can have a beneficial impact on the prevention of stroke. Despite the methodological shortcomings, compared with experimental designs, we think that our study also illustrates the merits of linking specific structural characteristics of practices to specific care processes as measured through an audit study. Because the audit study clearly demonstrated that there is room for improvement in the prevention of stroke, we recommend strengthening of policies towards practice organization in general practice. Meanwhile, it seems worthwhile to evaluate the quality of the preventive activities of general practitioners on a regular basis. The prevention of stroke as evaluated in our study can be a good model for this.

Acknowledgments

This study was supported by a grant from the Netherlands Organization for Scientific Research, The Hague, the Netherlands.

References

View Abstract