Priority Setting in General Practice: Health Priorities of Older Patients Differ from Treatment Priorities of Their Physicians

Population and recruitment

The study was conducted between September 2008 and January 2009. In the study design, a sample of 8 general practitioners (GP) and 32 patients (4 patients per GP) was defined. The recruitment of physicians took place in 5 pre-specified zip code areas (30 169, 30 171, 30 173, 30 952, 30 989) of Hannover and the surrounding region in Germany. Whereas 4 GPs were recruited through professional contacts of the authors, the others were respondents to cover letters with information leaflets sent to a random selection of 30 GPs (50% female) with their own practice in these zip code areas. The age of the GP, the size of the practice, and its location were not taken into account for recruitment. The participating physicians signed a written consent and agreed to provide practice characteristics and socio-demographic data. The study protocol was approved by the ethics committee of the Hannover Medical School (No. 5069).

Patients were recruited in the GPs’ practices on the basis of a pre-defined scheme determined by age and sex. Nurses were asked to recruit patients who came to the office for whatever reason and who consented to participate. They were advised to consecutively recruit the following 4 patients who presented at the office after 10 am on working days during a defined week: one male and one female patient aged 80 years or over, and one male and one female patient aged 70-79 years. If a patient with the desired age and sex could not be recruited on the first defined day, the recruitment was to be repeated on the following day. Inclusion criteria for patients were an age of 70 years or older and at least one contact with the physician in the previous 3 months. Exclusion criteria were long-term care dependency level II or III, known severe dementia, limited contractual capability or incapacity, insufficient language skills, severe hearing loss or deafness, current participation in another clinical trial, and no availability by telephone. Long-term care dependency levels II and III are characterized by a high level of physical, mental, and social vulnerability, which means that everyday life can be maintained only with outside help (1).

Participating patients were informed about the aims of the study and their tasks verbally and by means of an information leaflet; they then signed a written permission and provided socio-demographic data.

Assessment of health and everyday problems

The health and everyday problems of patients were assessed using the Standardized Assessment for Elderly Patients in Primary Care (STEP) (12). STEP is designed to uncover the complex health needs of older patients with multiple problems (12,13). It was revised as a German version in 2008, and has been tested and used in several trials (11,14-16). The German version consists of 76 single questions characterizing 44 health and everyday problems of older patients. For this study, we divided them into the following 10 health domains: functional condition and mobility (eg, problems with activities of daily living), 18 items; social environment and financial setting (eg, problems with housing conditions, no help in case of emergency), 6 items; somatic problems and medical examination (eg, pain, history of heart attack, hypertension), 23 items; mood (eg, depression, mourning), 4 items; lifestyle habits (eg, alcohol abuse, unhealthy nutrition), 5 items; immunization (flu, pneumonia, tetanus, diphtheria), 4 items; medication (eg, problems with number or intake of medication), 5 items; cognition/clock-drawing test (17), 1 item; foot abnormality (eg, pain, deformity), 5 items; and further problems (additional problems volunteered by the patient after the assessment).

A computer-aided program was developed to facilitate data collection using the STEP instrument. A study nurse completed the computer-aided STEP with each participating patient in the practice setting. The procedure took about 50 minutes. Patients received a list with an overview of the disclosed health and everyday problems.

Assessment and comparison of importance ratings

Patients and physicians independently rated the importance of each disclosed health problem. Whereas patients assessed only the importance for their everyday lives, physicians gave two importance ratings, one on the presumed importance for the patient's medical care and one on the importance for the patient's everyday life. Thus, each identified health problem received 3 importance ratings, 2 from the physician and 1 from the patient. Values were measured on an ordinal scale from 1 to 4: 1 – not at all important, 2 – slightly important, 3 – quite a bit important, and 4 – very important. The patients and their physicians discussed the STEP findings in a consultation.

To relate the importance ratings of patients and physicians, we compared a) the importance ratings of patients and the importance ratings of their physicians concerning their medical care and b) the importance ratings of patients and physicians concerning patients’ everyday lives. Ratings with values of 3 or 4 were considered as important and defined either as a health priority for the patient or a health/treatment priority for the physician. Values of 1 or 2 were not considered to be a priority.

Statistical analysis

Socio-demographic data of patients and physicians were analyzed descriptively. Based on the results of normal distribution tests, mean scores and standard deviation were calculated for the number of health problems per patient and for subgroups based on age, sex, and education. t test was used to determine the differences between age groups and sex, and variance analysis was used for education subgroups. Additional analyses were made for each single health domain. Since the number of problems in each health domain was not normally distributed, non-parametric statistics were used for calculations of centrality and dispersion (median, interquartile range), comparison of subgroups by age and sex (Mann-Whitney test), as well as education (Kruskal-Wallis test).

Importance ratings for each health problem were dichotomized: values of 1 and 2 were transformed to 1 – not important, and values of 3 and 4 were transformed to 2 – important. The observed agreement between importance ratings of patients and their physicians, as well as Cohen κ, a statistical measure of chance-corrected inter-rater agreement, were calculated. If raters are in complete agreement after allowing for chance, κ = 1. If there is no agreement among the raters other than what would be expected by chance, κ≤0. Values between 0 and 1 are interpreted in a range of agreement: 0.0-0.2, slight agreement; 0.21-0.4, fair agreement; 0.41-0.6, moderate agreement; 0.61-0.8, substantial agreement; and 0.81-1.0, nearly perfect agreement (18). Cohen κ was calculated for all problems and separately for each of the 10 health domains. P < 0.05 was considered significant. SPSS 17.0 (SPSS Inc., Chicago, IL, USA) was used for data analysis.

Descriptive analysis

Number of problems. In sum, 634 health problems were identified for 35 patients. Since the number of problems was normally distributed (Kolmogorov-Smirnov test, P = 0.105), parametric means were calculated. Each patient experienced a mean of 18.1 ± 9.2 health problems. Patients stated that most of their problems were in the somatic health domain, followed by problems in the “functional condition” and “immunization” domains. Subgroup analysis with t test and variance analyses for all problems showed that women had on average more problems than men (22.5 ± 10.2 vs 13.5 ± 5.1, P = 0.002). Patients with a low educational background reported more problems (23.3 ± 14.1) than patients with intermediate education (18.6 ± 7.1) or high education (10.3 ± 4.2) (P = 0.033). There was no significant difference in the number of problems between 16 patients aged 80 years and older (18.6 ± 9.1) and 19 patients aged 70-79 years (17.7 ± 9.6).

Non-parametric analyses per health domain are shown in Table 3. Women had more problems in the domains “functional condition,” “social environment,” “somatic problems,” and “mood.” In the “lifestyle-habits” domain, patients with intermediate education had more problems than patients with low or high educational background. In this domain, patients under 80 years had more problems than those 80 years and older.

Importance ratings. There were 537 of 634 problems (85%) that were rated by patients and physicians. Of these 537 health problems, 332 (62%) were considered to be important for patient's everyday life by patients and 334 (62%) by physicians. In addition, all physicians rated 294 of 537 problems (55%) to be important for patients' medical care.

When the health domains were ranked according to the frequency of important problems in each domain, patients considered “mood,” followed by “further problems” and “functional condition” to be most important. Physicians most frequently rated problems as important for health care in the “mood” domain, followed by “cognition” and “functional condition.” When physicians were asked to assign ratings from the patient's perspective, the domains of “mood,” “social environment,” and “functional condition” were most important (Table 4).

In order to compare the different perspectives of patients and physicians for problems in the different health domains, we analyzed the percentage of important problems per health domain separately for physicians and for patients. This means that for all problems in a health domain, patients’ importance ratings for their everyday life were compared with physicians’ importance ratings for medical care. Only differences in the relative frequency of important problems of 15% or more were considered. More problems were important to patients than to physicians in the health domains “mood,” “further problems,” and “functional conditions.” In contrast, more problems were important to physicians than to patients in the domains “cognition,” “medication,” and “immunization” (Table 5).

There were also differences among health domains in the percentage of problems rated as important by physicians for patients’ everyday lives and rated as important by patients. More problems were important to physicians than to patients in the domains “medication,” “social environment,” “lifestyle habits,” and “foot abnormality.” Compared with patients, physicians rated a smaller percentage of problems important in the health domain “further problems” (Table 6).

Proportional agreement and Cohen κ

Comparison of patients’ importance ratings for their everyday lives and physicians’ importance ratings for medical care. A total of 318 of 537 problems (59%) were identically rated by physicians and patients: 34% of problems were important to both groups and 25% were unimportant to both groups. Cohen κ was 0.18 (P < 0.001), which means that there is slight agreement between patients and physicians. The following domains showed at least moderate agreement on the importance between the two parties (Table 5): “social environment,” “foot abnormality,” and “cognition.”

Comparison of patients’ importance ratings for their everyday lives and physicians’ importance ratings for patients' everyday lives. A total of 334 out of 537 (62%) problems were rated identically by patients and physicians: 40% were important to both groups and 22% were unimportant to both groups. Cohen κ was 0.23 (P < 0.001), which indicates a fair agreement between patients and physicians. Fair agreement was also found for “foot abnormality,” “lifestyle habits,” and “immunization” (Table 6). No moderate agreements were observed.

Strengths and limitations of the study

Few research articles are available on health priorities of patients and the corresponding treatment priorities of their physicians. To our knowledge, this is the first study that examines individual patients’ and physicians’ priorities and their agreement in a general practice setting. In this study, health and treatment priorities are based on the existing health conditions. This is preferable to studies of hypothetical priority setting, which may be motivated by fear rather than by experience (10).

A limitation of this study is the small sample size (9 physicians, 35 patients). However, we used the disclosed health problems as the denominator in our analysis, which considerably enlarged the case number, leading to identification of a total of 634 problems, of which 537 (85%) were rated by patients and physicians. Nevertheless, the small sample size meant that it was not feasible to examine priorities for each of the 44 health problems separately and results have to be interpreted with caution.

A further limitation in the interpretation of results is the unknown validity and reliability of the importance ratings. Presently, we do not know whether patients and physicians will attribute the same importance to health problems when re-assessing them, for example, a month after starting treatment. If health care priorities are not stable, there is no benefit in prioritization. We are also aware that the given response options determine the outcome. A Likert scale with 4 response options and no middle value was chosen because we wanted the participants to make a clear choice. This may have not always represented the real evaluation of the participants. There are other ways of applying importance ratings, eg, ranking the most important problems or using a visual analogue scale. Future studies should assess different response scales and their retest reliability.

Implications for clinical practice

A structured and shared treatment planning process for older patients with multi-morbidity based on individual priorities may lead to a better care with mutual benefit for patients and physicians. This systematic and pro-active procedure of common priority setting does not seem intuitive and has to be learnt; physicians in general practices are not accustomed to such an approach (13). Shared priority setting requires organizational changes of consultations in general practices. First, this entails performing health overviews for a substantial proportion of older patients. Second, it requires systematic processing of various patients’ problems within one consultation. And third, it needs patient-centered communication to reduce treatable problems to an ideal minimum through shared priority setting. Presently, the short time schedule and the single-disease approach in consultations do not lend themselves to a holistic view about patients’ diverse problems.