Design and setting

A descriptive pilot study was performed in a primary healthcare centre belonging to the Swedish National Health Service in Region Västra Götaland. The centre serves approximately 10 000 patients and is staffed by about 10 physicians (five specialists in family medicine, three residents in family medicine, two licensed physicians without specialist competence and one intern). Two nursing homes, with approximately 100 residents, are attached to the centre.

Participants

All patients, ≥65 years of age, with a non-urgent consultation registered in the primary healthcare centre in 1 January to 31 January 2016, were included in the study. Thus, visits in the healthcare centre as well as visits in nursing homes and home visits were consecutively included if not caused by an urgent event. If a patient had >1 consultation during the study period, the first one, as well as the information available prior to this visit, was included in the study.

Data source

Data were retrospectively and manually extracted from electronic medical records (Asynja Visph and Journal III). The centre physicians were not informed beforehand that their work would be monitored the month in question. For each patient, socio-demographic and health-related data were retrieved including age, sex, residence, cognition, multi-dose drug dispensing (machine-dispensed unit bags with drugs that should be ingested concomitantly, for patients who have difficulties in handling their medications), drug treatment and morbidities. Regarding morbidities, we recorded the presence of medical conditions appearing in the Screening Tool of Older Persons’ potentially inappropriate prescriptions (STOPP), the Screening Tool to Alert to Right Treatment (START),19 the EU(7)-PIM list20 and the set of indicators of prescribing quality provided by the Swedish National Board of Health and Welfare.21

For each drug in the current medication list, at the end of the visit according to the medical records including electronic prescriptions with detailed information on prescribed products and doses, we recorded the substance name, the Anatomic Therapeutic Classification (ATC) code,22 and information as to whether the drug was prescribed for regular use or as needed. Drugs for external use were included only if having potential systemic effects. Combination drugs were counted as one drug. Two drugs including the same substance and formulation, for example, two different dosages of the same drug, were counted as one drug. The number of drugs that were used regularly (ie, not for a temporary condition such as an infection) was recorded.

We recorded if a procedure code, explicitly stating that a medication review had been performed, was present or absent in the medical records of the physician consultation in January 2016. If present, we recorded the actions taken and the content of the related documentation, based on the information available in the medical records and the electronic prescriptions. Thus, each drug in the medication list was categorised as unchanged, withdrawn, dosage adjusted, added or other. The changes in drug treatment were also summated at the patient level, categorised as ≥1 drugs being withdrawn, dosage adjusted, added, or other. Further, we recorded how the actions were documented in the medical records, focusing on whether the drug treatment problem was described, the goal with the treatment mentioned and planned follow-up expressed. In order to elucidate to what extent the procedure code medication review is documented at least annually, we extended the search in the medical record for this action, if not recorded in January 2016, to the entire year preceding the physician consultation. The contents of these medication reviews were not further analysed, as the focus of the study was drug treatment as of January 2016.

Assessments

Statistics

SPSS (IBM SPSS Statistics for Windows, V.19.0) was used for descriptive analyses concerning characteristics of patients and drug treatment by the absence/presence of ≥1 procedure codes for medication review over the last year. In the sample size considerations for this pilot study, we considered that including all patients fulfilling our inclusion criteria during 1 month, anticipated to amount to about a hundred, would yield useful results for the power calculations in the full-scale study, that is, reasonably certain prevalence figures on suboptimal drug treatment as well as information on the prevalence of documented medication reviews. Values are presented as numbers (percentages) as well as mean±SD or median and IQR or range where appropriate. As the Swedish regulations declare that physicians shall perform annual medication reviews for persons ≥75 years of age with ≥5 drugs in the medication list, these categories were also used.

Patient and public involvement

This research was done without patient involvement. Patients were not invited to comment on the study design and were not consulted to develop patient relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

Main findings

In this study, we show that more than 6 in 10 patients in primary care had appropriate drug treatment, when assessed from an overall perspective by a physician, taking morbidity, life expectancy and concurrent drug treatment into account. Only 1 in 10 patients had suboptimal drug treatment where an immediate change was suggested to be prioritised, mostly concerning a withdrawal of sedatives/anxiolytics. For one in four patients, the suboptimal treatment was suggested to be of lower priority, predominantly concerning the withdrawal of a PPI.

According to the procedure codes in the medical records, primary care physicians documented medication reviews in about half of all older people at least annually, with focus on those of old age with extensive comorbidities and long medication lists. For one in three patients, the medication review resulted in an active change of treatment, most often a dosage adjustment as, for example, increasing the metformin dose due to high glycated haemoglobin.

Strengths and limitations

To the best of our knowledge this is the first study to provide information on the quality of drug treatment in older people in primary care from a broad family physician perspective, taking individual factors and priorities in the complex clinical situation into account. Thereby, our pilot study contributes valuable information for sample size consideration in future studies on prescribing practices and interventions for improved drug treatment. Another strength is that the physician assessment was preceded by the application of numerous screening tools for PIMs and PPOs, thus minimising the risk of overlooking important aspects regarding the prescribing quality. Indeed, only 1 in 20 patients did not have any PIMs/PPOs, indicating that our screening tools captured potentially suboptimal treatment to a greater extent than many other studies.4 5 17 An additional strength is the fact that only one patient was excluded, an important aspect for generalisability.

Limitations of this study include the sample size and the fact that data were obtained from one healthcare centre only. This may reduce the external validity as centres may differ from each other regarding, for instance, population covered and prescribing traditions.24 However, a limited sample size may be justified in a pilot study; a full-scale study, sized on the basis of these pilot results, has been initiated. Another limitation is that the physician assessments were performed by one person and not a team. However, this physician had relevant expertise as well as previous experience of scientifically assessing pharmacotherapy in older people.23 In addition, ambiguities in the assessments were discussed with a senior clinical pharmacologist. Nevertheless, full-scale studies should preferably include two assessors and consensus discussions to provide information on inter-rater agreement. Another limitation is that the study was restricted to information in the medical records from primary care. Indeed, the assessments were based on available data in routine care, carrying a risk that the information about the medical history, diagnoses, examinations and laboratory tests had not been documented or appropriately coded in the medical records. Yet, this data source is more comprehensive than drug registers, either alone or linked to other registers, which are frequently used in studies on prescribing practices.25

Interpretation

The apparent discrepancy between the prevalence of suboptimal drug treatment according to the physician assessment and the screening tools illustrates the importance of taking future research on prescribing practices a step further to increase the medical relevance. Our results suggest that the limitations regarding the concurrent validity of general screening tools in hip fracture patients and the very old18 26–28 may also apply to older people in general in primary care. Indeed, although the positive predictive value for several indicator sets has been reported to be acceptable when determined separately,26 27 the PIMs/PPOs may not be relevant when taking the overall clinical picture into account, that is, from a family physician perspective.

The low prevalence of suboptimal drug treatment where an immediate change was suggested contributes to the understanding of the medical practice in primary care. In fact, the level of priority, according to our definition, may represent a balance between feasibility and expected patient harm; changes that can be achieved with reasonable efforts in relation to the anticipated patient benefit will get a higher priority than the opposite. Indeed, the changes suggested by the assessing physician predominantly concerned sedatives/anxiolytics and PPIs, withdrawals of which may not easily be achieved once they have been initiated.24 29 The high prevalence of PPI without an underlying indication is in agreement with a recent population-based register-based study where no underlying disease-related or drug-related reason could be identified for 4 in 10 patients on long-term treatment with PPI.30 Our pilot findings also contributes to the understanding of the limited implementation of recommendations for drug changes after medication reviews by third parties.31

Interestingly, the proportion of patients with suboptimal drug treatment, irrespective of priority, was similar for patients with and without a procedure code explicitly stating that a medication review had been performed over the last year. However, patients with such a procedure code had a greater burden of disease, also illustrated by the higher number of drugs in the medication list.32 Thus, the treatment in multimorbid patients can be expected to be more complex, and it is reassuring that drug treatment quality is acceptable to the same extent also in these patients. Although this pilot study does not allow conclusions, one may speculate that performed medication reviews may have contributed to appropriate drug treatment in multimorbid patients requiring complex treatment considerations.

Patients as defined in the Swedish regulations to receive a medication review at least annually, that is, patients ≥75 years of age with ≥5 drugs in the medication list, had this procedure code documented to a large extent. The adherence to this regulation, despite the absence of an established association between age and drug treatment quality,18 and the limitations of cut-offs of number of drugs in the medication list to reflect quality of drug treatment,33 may illustrate that monetary compensations can direct healthcare efforts to some extent. However, the evidence for effects of pay for performance policies in improving quality of care and health outcomes is limited.34 Indeed, a recent study, investigating the impact of pay for performance regarding the procedure code ‘medication review’ found no apparent association between this code and the prevalence of PIMs, although the number of patients with such a code increased.35

The actions performed within the documented medication reviews were quite few, as only 1 out of 14 drugs were acted on. This finding may give the impression that medication reviews by the attending physician is a minor effort. However, as a medication review implies a medical assessment, ascertaining that the drug treatment is reasonable given the current health status, the decision not to change a drug may be just as important. Moreover, the basic part of a medication review, the medication reconciliation, was not captured in the present study.

We were happy to find that the problem underlying the treatment action was documented in the medical records in all but one case. Indeed, such information is important in transitions of care. Nevertheless, our results also suggest that documentation is an area with potential for improvement; the goal of the drug treatment change was only stated in about half of the cases, and the planned follow-up in a minority.

In conclusion, the results of this pilot study suggest that current prescribing practices in older people in primary care is in general appropriate, especially when medical priorities are considered. Efforts for rational use of medicines over the last decades may have contributed to this finding, including easily available information on recommended drugs,36 37 computerised decision support,38 39 and indicators of prescribing quality for benchmarking and educational purposes.21 40 Indeed, although the latter may reflect quality of drug treatment to a limited extent, they may be valuable in particular for junior physicians, as they reflect expert opinions on current best practice for an average older patient. We also want to emphasise that, although our findings indicate that drug treatment in older people in primary care may be quite good, endeavours for improvements are always desirable. Our results, reflecting drug treatment quality from a broad and clinically relevant perspective taking individual factors and medical priorities in the complex clinical situation into account, may be useful in sample size considerations for future non-randomised and randomised studies, focusing to factors associated with prescribing practices and interventions.