Prognostic value of orthogeriatric assessment parameters on mortality: a 2-year follow-up

Typical geriatric assessment parameters were chosen in the determination of functional status and comorbidities. The Parker-Mobility Score (PMS) prior to the debilitating trauma was achieved and patients were split into three groups: PMS 0–3, PMS 4–6 and PMS 7–9.

ADL were assessed by Barthel Index (BI) after completion of the stationary treatment right before discharge, here again three groups were formed: BI 0–30, BI 35–65 and BI 70–100.

The comorbidities were determined in the medical history on admission and the respective CCI was calculated, dividing patients in the groups CCI 0–1, CCI 2–3 and CCI ≥ 4.

Cognitive status was assessed on admission either by known dementia or by Mini-Mental-Status Examination (MMSE). As literature suggests, a cutoff of 24 score points was used to subdivide into demented and not demented, labelling every result with less than 24 score points as demented and higher results as not demented [14]. The geriatric depression scale (GDS) was used to display the distribution of prevailing depression, setting a cutoff point of 5. Every result with at most 5 points labelled as not depressed, every result higher as depressed [15].

Sarcopenia on admission was determined by the patients’ respective maximal calf circumference measurements using a cutoff value of 33 cm [16]. Every calf-circumference being at least 33 cm labelled the patient as not sarcopenic, vice versa as sarcopenic with the calf-circumference being lower.

Falls frequency was assessed by the medical history, if there were at least two falls in 6 months prior to admission the according patient was grouped as a frequent faller.

The previous POR was determined on every admission, discriminating between private home (PV), assisted living (AL) and nursing home (NH).

The existing CL was recorded and used as an indirect indicator of the need for care and reduced ADL. Depending on the national provisions during the assessment period, there were 3 CL’s splitting patients into four groups: no CL at all, CL 1, CL 2 and CL 3. The categorization to the respective CL was dependent on the frequency of daily support needed and daily hours spent for care by caregivers [17].

The LOS was calculated and is represented in the three groups LOS < 1 week, LOS 1–2 weeks and LOS > 2 weeks. LOS itself was examined for its dependency on the respective discharge accommodation of previously home-dwelling patients.

Orthogeriatric co-management was provided for every patient. Next to being treated by both an orthopedic and geriatric specialist, patients attended daily physio- (twice) and ergotherapy. Fractures of the lower limb were generally treated surgically to provide early mobilisation. Humeral and forearm fractures were mostly addressed operatively (87% of cases of follow-up cohort). Vertebral fractures underwent a decisional algorithm taking into account fracture morphology, clinical course, pain and course of radiographic changes resulting in a fifty–fifty ratio of surgical to conservative treatment.

Pathologic fractures were treated according to the patient’s respective prognosis either by internal fixation or prosthesis and if possible, with combined radiation therapy.

Soft tissue infections were treated by incision and debridement, prosthetic infections were treated by revision of prosthesis and either total or partial exchange depending on infection’s severity and acuity.

Pelvic and rib fractures were treated conservatively.

Follow-up was generated after 2 years by sending questionnaires to patients and relatives determining survival. In case of receiving no response by mail, a maximum of 5 attempts of contact via phone call was performed. Should the respective patient having been deceased, the exact month of death was inquired. Patients that were lost to follow up were excluded.

SPSS (IBM Corp., Armonk, New York, USA) was used for data analysis. Mean values and standard deviations were determined for continuous variables. T tests for independent samples were used for comparison of mean values. Mortality differences between the variable groups were evaluated by Fisher’s exact test and log-rank test for their significance. A Cox-proportional model was set up to adjust the according mortality risks of assessment parameters in a uni- and multivariate analyses. To provide a valid analysis at least ten events per independent variable had to be obtained [18].

Comparing age groups 71–80, 81–90 and 91–95 years in both 1- and 2-year mortality a linear increase in mortality could be observed (p < 0.001). Gender shows a lower 1-year mortality for female patients (p = 0.015). This difference vanishes after 2 years showing death rates of 41.8% for men and of 37.3% for women (p = 0.31).

PMS and BI display inverse linear correlations between mobility, respective function and mortality (Fig. 1). The CCI also shows stepwise reduced survival with more existing comorbidities. Every association shows statistical significance (p < 0.05). 2-year mortality in the respective best and weakest groups is as following for PMS: 18.3%: 53.4%, BI: 12.7%: 54.9% and CCI: 26.8%: 46.0%.

Preexisting dementia significantly correlated with increased mortality. 2-year mortality was 31.8% for cognitively capable patients, 51.5% for patients with dementia (p = 0.000). Existing depression as screened by GDS showed no significant influence on mortality (2-year mortality: 31.9% for GDS ≤ 5, 38.3% for GDS > 5, p = 0.176).

Sarcopenia showed a significant relation with increased mortality, displaying a 2-year mortality of 45.9% compared to 29.0% for non-sarcopenic patients (p = 0.000). Frequent fallers had lower survival rates in comparison to non-frequent fallers, showing significance after 2 years of observation (mortality rates after 2 years: 27.2%: 42.1%, p = 0.001).

No association between mortality and LOS was seen in the cohort. Patients who were discharged in the first week showed a tendency to the highest death rates after 1 year (33.3%) and after 2 years (42.0%). LOS was dependent on the respective discharge accommodation. It was significantly raised in patients that lived at home and were discharged to rehabilitation (18.4 days; n = 207; p < 0.001) or to a nursing home (14.8 days; n = 51; p = 0.048) compared to patients that were discharged to their homes (12 days; n = 193).

A multivariate Cox-regression was performed being displayed in Table 5. Both GDS and LOS were removed from the model, not showing significant impact in the performed univariate analysis. After correction 381 cases (130 events) with overall complete assessments in the ten involved risk factors remained. The strongest independent correlation demonstrated PMS with a stepwise HR of 1.81 (95%CI: 1.373–2.397) for more immobile patients. BI showed a significant HR of 1.64 (95%CI: 1.180–2.290) for cohorts capable of less ADL. Age also persisted as an independent factor displaying a HR of 1.04 (95%CI: 1.004–1.067) per year. Every other parameter did not remain as a significant influential factor. Higher CL, more institutionalized POR and frequent falling even insignificantly correlated with improved survival after correction for cofactors.

We could examine a sufficient case number of orthogeriatric co-managed patients to provide a valid statistical evaluation. Important geriatric assessment parameters were included. Nevertheless, not every possible variable that could have affected mortality was considered, as especially in-hospital complications were missing in our analysis. These are known to have significant relevance for orthogeriatric patients’ survival [33]. Here an actual detailed analysis, which would go beyond the scope of this article, is in publication. Literature often makes use of the ASA Score which is not integrated in the actual investigation. The patients’ muscle strength could also not be taken into account as another predictive parameter due to missing of appropriate measurements. We did not consider different fracture types and the respective predictive importance of each parameter. A stratification of the ten variables included in the analysis could not be performed fracturewise due to too few remaining events in the respective fracture groups. A more rudimentary fracturewise stratification was performed in a previous publication. We, therefore, again have to emphasize our analysis as being an overview including a mean orthogeriatric patient cohort suffering from different injuries and causes for admission. It has to be taken into account that some injuries impact mortality more heavily (e.g., hip fractures) than others (e.g., forearm fractures) [34]. The group POR, LOS and GDS were excluded in the multivariate analysis, for not having shown univariate significance. Although this was also the case for gender, it was retained in the analysis for this factor to be known to have some impact on mortality considering the existing literature. To be able to deliver a valid multivariate analysis, ten events per variate included to the analysis are required to obtain valid results [18]. Because of the lack of data in several assessment parameters the number of cases that could be included into the Cox-regression analysis was reduced to 381 including 130 events. This still granted at least 10 events per involved variable. Still there was a risk of biasing by having excluded 280 cases. The diversity of the observed cohorts, the different end points (30-day, 1-year, 5-year mortality etc.) and the variety of parameters that are included in the analyses complicate comparisons to literature. Even though there are consensual recommendations for orthogeriatric outcome parameters, general and consistent adoption may be difficult in practice.