Municipal resources and patient outcomes through the first year after a hip fracture

We conducted a nationwide cohort study on hip fractures occurring in the period from 1 January 2011 to 31 December 2012, based on data from the Norwegian Hip Fracture Register (NHFR) [14], with 1-year follow-up. The cohort was merged with the Norwegian Patient Register (NPR) at the Norwegian Directorate of Health [15] by personal identification, and municipality resources were incorporated from the Municipality-State-Reporting (KOSTRA) at Statistics Norway [16].

The Norwegian Hip Fracture Register contains information on patients operated for hip fracture from 2005 at all hospitals in Norway that perform these operations [14]. All hip fractures reported to NHFR in the study period were included. Variables extracted for this study comprised the date of surgery, patients’ age and gender, municipality of residence, preoperative ASA (American Society of Anesthesiologists) score, health-related quality of life (EQ-5D-3 L) preoperative (reported 4 month postoperative) and at 12 months postoperative, and in case of death within 1 year after operation, the date of death. The EQ-5D-3 L is a descriptive tool assessing 5 dimensions: the level of mobility, self-care, usual activity, pain/discomfort, and anxiety/depression [17]. Each question has 3 levels: no problems, some problems, and severe problems. A visual analog scale (VAS) described how satisfied the patients were with their general health condition. The value 0 represented very dissatisfied while the value 100 represented very satisfied.

NHFR data was merged with data in the Norwegian Patient Register (NPR) using the unique 11-digit personal identity number assigned to all residents of Norway. NPR is a national health register containing information on all patients who have received specialist health services from 1996 [15]. We extracted data from NPR regarding length of hospital stay for primary hip fracture surgery (date of admission and discharge), and hospital readmissions during the first year post surgery due to complications such as bleeding, wound infection, mechanical problem, or venous thromboembolic complications (ICD-10 diagnostic codes T81 and T84; date of admission and discharge) [18].

Municipality-State-Reporting (KOSTRA) is a national information system that contains management information on municipal key activities including health services from 1995 [16]. Administrative data (e.g. patients, staffing and costs) are collected quarterly from all 428 municipalities. We extracted data on municipalities’ centrality (i.e. location of municipalities in relation to urban settlements) grouped in four categories (least central, less central, quite central and central). Municipality age profile was defined by two variables; percent population above 67 years and percent population above 80 years. KOSTRA contains no data regarding the use of municipal rehabilitation services. We therefore decided to include indirect measures, i.e. available healthcare resources in the municipality relevant for rehabilitation, provided > 80% completeness of data; overall staffing (fulltime positions in rehabilitation per 10,000 inhabitants, i.e. home based and nursing home services); nursing home resources were expressed by number of short-term stays divided by number of inhabitants in the municipality, and staffing (available doctor and physiotherapist hour/nursing home patient/week). All variables were divided in quartiles based on the study population.

The study population comprised all patients operated due to a hip fracture and reported to the NHFR between 1 January 2011 and 31 December 2012. Follow-up lasted for 1 year after the hip fracture or until death.

Patient characteristics were given as percentages and for the continuous variable as mean, standard deviation (SD), median as well as minimum and maximum values. The resources in the 428 municipalities were listed as percentages, mean and SD. To quantify the association between the age structure of the population and the different municipality resources Pearson bivariate correlation coefficient was calculated.

Mortality was analyzed both as death within 30 days, and death within 1 year after hip fracture surgery. Cox regression model was used to estimate relative risk (RR) and 95% confidence interval (CI) for mortality in different levels of municipality resources with the lowest quartile as reference category. The models were adjusted for patients’ age, sex, ASA, and the municipalities’ centrality.

Each dimension in EQ-5D-3 L was analyzed separately and was dichotomized with the two most favorable levels (no or some problem) as reference, to which the poorest outcome was compared (mobility: bedridden; self-care: unable to wash or dress; usual activity: unable to perform; pain/discomfort: extreme; anxiety/depression: severe). To account for the large proportion of death within 1 year after surgery when analyzing EQ-5D-3 L, invers probability weight (IPW) was estimated from a logistic regression model including variables associated with mortality. To estimate the effect of municipality resources on health-related quality of life, logistic regression analysis with IPW was applied. The estimated odds ratio (OR) was adjusted for age, sex, ASA and centrality. To investigate whether the municipality could explain some of the variation in EQ-5D-3 L or mortality, we performed mixed effect logistic regression with municipality as random effect. Since the intraclass correlation coefficient was in the magnitude of 0.03 and lower, ordinary logistic and cox regression was maintained.

The study population comprised 15,757 patients (68.6% women) with a mean age of 80.8 years (SD 11.6 years), and a median ASA score of 3. The median length of index hospital stay was 6 days. On average, patients were readmitted 1.5 (SD 1.2) times during the first year postoperatively; 8.4% of the patients for specific complications after hip fracture surgery, Table 1.

There were 428 municipalities in Norway in the study period; 35% being characterized as least central, while less, quite and most central were 12, 18 and 35% respectively. Average available doctor time in nursing home increased from 2011 to 2012 by 2.4 min/patient/week (p < 0.002). The results showed no differences in available physiotherapist time in nursing home, mean numbers of short-term stays in nursing home, or overall municipal staff time regarding rehabilitation, illustrated in Table 2. The correlations between the age structure of the population and municipality resources were in the magnitude of 0.24 or less.

All-cause mortality was 8.6% within 30 days, and 25.3% within 12 months postoperatively, Table 1. Mortality was lower in municipalities with the highest overall staff time regarding rehabilitation and marginally higher in municipalities with available nursing home doctor time within the third quartile. Physiotherapist time and number of short-term stays in nursing home was not associated with mortality, Table 3.

Altogether 51% of patients still alive 12 months after hip fracture surgery completed the EQ-5D-3 L questionnaire. The percentage of the population reporting severe problems in EQ-5D-3 L was low before the fracture. After 12 months, a larger percentage was bedridden, unable to wash/dress, and/or unable to perform usual activity; while extreme pain/discomfort and severe anxiety/depression were at about pre-fracture level. An additional figure shows this in more detail [see Additional file 1: Figure S1]. In municipalities with the highest percentage of the population above 80 years, there were lower rates of self-reported anxiety/depression 12 months after surgery, and the estimated mean EQ-VAS score for general health was 3.2 (95% CI: 0.6–5.8) higher compared to the lowest quartile, Table 4. Also in the municipalities with the highest quartile percentage population above 67 years, patients had a higher EQ-VAS score compared to the group with the lowest percentages (3.4; 95% CI: 0.6–6.2), Table 4. Outcomes in municipalities with the highest staffing quotient did not differ from those with the lowest staffing, Table 4. There was no significant association between EQ-VAS and nursing homes resources.

This study showed only a minor impact of municipal resources on survival and quality of life through the first year after a hip fracture. Mortality was lower in municipalities with most overall staff time regarding rehabilitation, but marginally higher in municipalities with relatively more available nursing home doctor time. Health-related quality of life, in terms of overall satisfaction and five distinct dimensions, was not associated with municipal resources in the studied sample. A high percentage of the population above 80 years, was associated with lower rates of self-reported anxiety/depression 12 months after surgery, as well as higher scores for general health.

Whereas the large sample size is the major strength of our study, the sparse amount of information available per individual is the most important limitation. In this nationwide study, data from two quality health registries provided valid and reliable measures for demographics, hip fracture, hospital readmission and death. The NHFR comprises 90% of all hip fracture operations in Norway [19]; we assume that underreporting is non-differential, with no impact on the results. Data on health-related quality of life was available for half of the population alive after 12 months; thus, we cannot rule out reporting bias in either direction. At present, there is a lack of reliable municipality-based clinical data at the patient level (e.g. diagnoses, functional ability), and contribution of health personnel (e.g. profession, competence, care level). Data from Municipal-State-reporting of community services relevant for rehabilitation after hip fracture are the best proxy measures available; however, they represent crude measures and sources of error.

Hospital based data of factors such as fracture type, mode of operation and inpatient treatment may influence patients’ potential for rehabilitation but were not included in the present study. The largest municipalities (high centrality) in Norway are hosting several hospitals, while several smaller municipalities (lower centrality) are covered by one hospital. Mixed effect logistic regression with municipality as random effect revealed that municipality could not explain the variation in EQ-5D-3 L or mortality.

Both organizational factors in the health care system (mainly public in Norway), and available rehabilitation options must be considered in transferring our results to other countries.