Rate and predictors for non-attendance of patients undergoing hospital outpatient treatment for chronic diseases: a register-based cohort study

This study is a historic register-based cohort study. We included patients aged 18 years or older who were undergoing outpatient treatment for one or more selected common chronic diseases.

The common chronic diseases were chosen according to those defined as major by the Danish Health Authority and frequently encountered in hospital outpatient settings [26‐28].

We included patients with type 1 diabetes mellitus (corresponding ICD–10 code: E10), type 2 diabetes mellitus (E11), heart failure (I50), other chronic obstructive pulmonary disease (COPD) (J44), asthma (J45), rheumatoid arthritis (M05), and osteoporosis (M80 and M81). Patients had to be in ongoing treatment on July 1, 2013 at Hospital Lillebaelt (Denmark).

Hospital Lillebaelt is a regional hospital with a rural and urban catchment population of nearly 300,000 inhabitants [29]. Information on patient appointments was extracted for the period from July 1, 2013 to June 30, 2015.

The Danish National Health Service provides access to tax-financed health care for Denmark’s 5.8 million citizens [30]. All Danish hospitals, public and private, report inpatient and outpatient contacts to the National Patient Register (NPR), which holds longitudinal data derived from the electronic medical records of all inpatients of Danish hospitals since 1977 and outpatients since 1994. NPR contains discharge diagnosis codes according to ICD-8 from 1977 to 1993 and adhering to ICD-10 from 1994 onwards [31]. Since 2000, it has also served as the basis for reimbursement via the Diagnostic Related Group system [32]. NPR comprises information on appointments that are actualized in attendance. However, NPR does not hold information on non-attendance as hospitals are not reimbursed for non-attended appointments. NPR was used to extract the study population and to obtain information on the patients’ age and gender as well as on all outpatient appointments during the study period.

As non-attendance is not registered in the NPR, we used information from Hospital Lillebaelt, where non-attendance is documented in the electronic medical records. Occurrences of non-attendance are registered on a daily basis, as health providers construct a schedule based on whether patients have attended, canceled, rebooked, or have not attended appointments. Non-attendance is not registered for the first appointment of the hospital outpatient treatment course.

Further, we gathered information on patients’ marital status from the Danish Civil Registration System [33], on their educational levels from the Danish Education registers [34], and on their occupational status and income from Danish registers on personal income and transfer payments at Statistics Denmark [35]. The links between data from different data sources are made possible using the unique 10-digit identification number assigned to all Danish citizens at birth or at first immigration [33].

The outcome variable was non-attendance. One appointment with or without attendance was included daily. If patients had two appointments on the same day with one registered as attendance and one registered as non-attendance, we categorized this day as attendance, as the non-attendance may have been caused by the hospital; for example, by a prolonged visit in one clinic or by a reservation of for medical imaging that was not used.

Explanatory variables were chosen based on their influence on non-attendance according to previous studies [5]. Further, we included the variables “specific chronic disease” and “number of outpatient treatment courses” as we wanted to know the influence of disease and treatment characteristics. Information on the following explanatory variables was extracted:

The study did not involve any direct contact with patients or access to patient files. Thus, informed consent or ethical approval were not required [36]. The study was registered with the Danish Data Protection Agency (J.nr. 18/35695).

The included patients had a total of 82,898 appointments during the study period, of which 5% (4393 appointments) were not attended (see Fig. 1: Flowchart of patient inclusion). A total of 195 appointments had occurrences of attendance and non-attendance on the same day; these were categorized as attendances.

The characteristics of patients who attended appointments and those who did not attend are shown in Table 1. There were significant differences in nearly all covariates between attendees and non-attendees. Those who did not attend appointments were more often young patients (40% vs. 19% aged 18–49 p < 0.001), male (58% vs. 52% p < 0.001), unmarried (53% vs. 42% p < 0.001) and low educated (44% vs. 41% with only primary school education p < 0.001). Retired patients had the lowest non-attendance rate of the occupational groups (31% vs. 53%). Among the specific disease groups, patients with type 1 diabetes (28% vs. 20% p < 0.001) had a higher non-attendance rate while a lower non-attendance rate was seen among patients with rheumatoid arthritis (11% vs. 14% p < 0.001) and osteoporosis (4% vs. 8% p < 0.001).

The results of the multivariate logistic regression are shown in Table 2. In our primary multivariate logistic model (a) young patients (aged 18 ≤ 29 years) had the highest odds of missing their appointments (odds ratio (OR) 4.17; 95% confidence interval (CI) 2.70–6.42). Other significant predictors included middle-younger age (OR 2.58; CI 1.76–3.78 for 30 ≤ 39 years; OR 2.09; CI 1.47–2.98 for 40 ≤ 49 years), the male gender (OR 1.35; CI 1.20–1.52), being unmarried (OR 1.39; CI 1.24–1.56), having a low educational level (OR 1.18; CI 1.01–1.40), and being on long-term welfare payments (OR 1.48; CI 1.13–1.93). There were no specific diseases associated with higher odds of missing appointments.

The secondary multivariate logistic regression (b) showed that patients undertaking several outpatient treatment courses were not significantly different from patients undertaking one outpatient treatment course.

Our study has important strengths. First, it included all patients undergoing hospital outpatient treatment for the selected diseases at Hospital Lillebaelt. Second – information from other registries for all patients was available because of the unique identification number. Third, the diagnoses in NPR are generally of high validity [31]. Of the diagnoses we included, only osteoporosis has not been validated in a hospital outpatient setting. The positive predictive value (PPV) for each of the other diagnoses is as follows: type 1 diabetes and type 2 diabetes (PPV = 96.0 (86.5–98.9)), heart failure (PPV = 100 (92.9–100)), COPD (PPV = 100 (92.9–100)), asthma (PPV = 65.3 (62.2–68.3)), and rheumatoid arthritis (PPV = 98.0 (89.5–100)) [31]. Third, we had a large study sample including patients from different hospital clinics.

However, there are also some limitations to our study. First, it is based on administrative data. As all attended appointments are financially reimbursed by the government, we presume that the data are complete. In contrast, appointments without attendance are not financially reimbursed, and although there is a policy at Hospital Lillebaelt of documenting non-attendance, we cannot be sure, that it is always registered correctly. A limitation is also the possibility of non-attendance being caused by hospital admission, though admissions are often known in the documenting process of non-attendance. Further, the database of Lillebaelt does not record missed appointments that would have provided the first contact for a new outpatient treatment course. This possibly introduced a component of measurement bias to the study, resulting in the under- or overestimation of the reported rate. However, the predictors would presumably have remained unaffected, as the bias would have been equally distributed within the groups of attendees and non-attendees, thereby retaining their proportions. Additionally, we chose to categorize information on patients who attended an appointment and missed an appointment on the same day under attended appointments. As this was the case for only 195 appointments, it could not have biased the results. Nonetheless, a limitation can be found in that the collection of information on predictors was extracted on July 1, 2013, thereby earlier than the actual appointments. This may have resulted in an under- or overestimation of the “number of outpatient treatment courses.” We presume that variations in the other predictors have not greatly affected the overall results as they are unlikely to have varied significantly over the two-year study period.

While we set out to investigate the effect of having multiple chronic diseases on the rate of non-attendance, we only investigated diseases treatment at hospital outpatient level. Patients may have other chronic diseases that they undergo treatment for in primary care. Therefore, we can only investigate the effect of having multiple chronic diseases at a hospital outpatient level.

In this study we investigated factors we could extract from the administrative system which do not contain information on patient reported outcome measures. Information regarding the patient’s perception could be investigated by means of qualitative interviews. We did not investigate if visits were in the form of face-to-face meeting, telephone consultations or other kind of telehealth communication which could have been relevant for subgroup analysis. Furthermore, it was not possible to divide appointments based on their purpose, e.g., acute appointment; evaluate a medication dosage change; routine follow-up to monitor chronic conditions; etc. which is a limitation.

Finally, hospital Lillebaelt has a rural and urban catchment population that covers 5% of the Danish population, but other Danish hospitals may differ in catchment population. However, overall we find our conclusions applicable to other Danish hospitals and health-care systems like the Danish.

This study showed that Danish patients with chronic diseases miss one in every 20 appointments (5%). A major part of hospital outpatient activity involves patients in treatment for chronic diseases and non-attendance by members of this group is very inefficient for the hospital. Further, perhaps especially for patients with chronic conditions, continuity of care is an important aspect and non-attendance interrupts this continuity. Therefore, we argue that a further reduction in the non-attendance rate is important.

The results from our study on chronic patients in hospital outpatient clinics can be made operational in a number of ways. Knowledge about predictors for potential non-attenders can assist in targeted interventions; for example, studies on outpatient reminder systems have shown that they reduce non-attendance [12, 37]. In a hospital were such a system has been implemented it is relevant to investigate to what extent this affects non-attendance rates among chronic patient groups.

In this study we investigated the extent of non-attendance for patients in long-term treatment and some factors for non-attendance. Information regarding mechanisms on why these patients do not attend their appointments by qualitative methods should be obtained e.g., if non-attendance is caused by forgetfulness a reminder system could help, but patients may miss appointments for other reasons that are important to discover if we wish to find better ways to organize their treatment courses.

A final note on the clinical implication of this study regards the clinical relevance of the prognostic factors. We find our results robust because of the relatively large sample size. It is however for the interested clinic or hospital management to decide which prognostic factors are of clinical relevance which will depend on factors such as context and population size.