Short-term morbidity factors associated with length of hospital stay (LOS): development and validation of a Hip Fracture specific postoperative morbidity survey (HF-POMS).

Background: We aimed to describe and quantify postoperative complications in the older hip fracture population, develop and validate a hip fracture postoperative morbidity survey tool (HF–POMS). Methods: A prospective clinical observation study of patients (≥ 70 years) admitted for emergency hip fracture surgery, was conducted across three English National Health Service hospitals. Outcome data items were developed from the Postoperative Morbidity Survey (POMS), Cardiac-POMS, hip fracture postoperative literature and orthogeriatric clinical team input. Postoperative outcome data were collected on days 1, 3, 5, 8 and 15; 341 patients participated. Results: A 12-domain HF-POMS tool was developed with acceptable construct validity on all HF–POMS days. Patients with high perioperative risk scores as measured by the NHFS and ASA grade were more prone to develop HF–POMS defined morbidities. High morbidity rates occurred in the following domains; renal, ambulation assistance, pain and infectious. Presence of any morbidity on postoperative days 8 and 15 was associated with subsequent length of stay of 3.08 days (95% CI 0.90 – 5.26, p= 0.005) and 15.81 days (95% CI 13.35 – 18.27, p = 0.001) respectively. Observed average length of stay was 16.9 days. HF–POMS is a reliable and valid tool for measuring early postoperative complications in hip fracture patients. Additional domains are necessary to account for all morbidity aspects in this patient population compared to the original POMS. Conclusion: Many patients remained in hospital for non-medical reasons. HF-POMS may be a useful tool to assist in discharge planning and randomised control trial outcome definitions.


Introduction
Elderly patients undergoing emergency hip fracture surgery have a high risk of perioperative complications and death. In the UK, postoperative mortality ranges between 7-11% at one month, 16-28% at six months and 22-37% at one year 1 2 . In-hospital postoperative morbidities are more frequent (17-50%) and often complex. A significant proportion of hip fracture survivors have decreased ability to perform activities of daily living; 50% do not regain their pre-fracture functional status and 10-20% of those admitted from home require long term institutional care 2 . Accurate identification, quantification and care of short-term morbidity after hip fracture surgery is important since these conditions are strongly associated with mortality and length of hospital stay 3 , impact long-term survival 4 , and resource management.
Current hip fracture care lacks a standardised categorisation of clearly defined early undesirable postoperative clinical outcome measures that can individually or collectively result in lower quality of life. The postoperative morbidity survey (POMS, appendix 1) 5 was developed from a heterogeneous elective patient population producing a nine-domain tool, each domain with specific defined criteria, to provide consensus in measurement and reporting of outcome following surgery. Its primary purpose is to identify the presence of complications that delay hospital discharge. The tool was further validated in several studies 3 6-8 demonstrating acceptable inter-rater reliability, accuracy in capturing criteria defined morbidities, and association with length of hospital stay (LOS). Other studies have shown POMS to be associated with both LOS and other pre-operative comorbidities [9][10][11] . However, to date the POMS has not been evaluated specifically in hip fracture patients. Its use and performance is likely to vary when applied to specific patient populations, reflected in the modification of the original POMS from a nine-domain item to the cardiac postoperative morbidity score (C-POMS, appendix1) 12 which is a 13-domain discharge 4 score. The distinctive morbidity aspects and disparities across surgical specialities necessitate the need for a speciality specific POMS 7 12 . We hypothesise that patients undergoing fragility hip fracture surgery are most likely to express morbidity related to mobility, and pain aspects due to both mechanical injury and surgical trauma 13 14 . After hip fracture from a fall, there may be psychological implications, fear and anxiety about walking again. The aims of this study were to describe and quantify postoperative complications in the older hip fracture population, and develop and validate a hip fracture postoperative morbidity survey tool (HF-POMS). Patients were recruited from orthopaedic and trauma wards following surgery. Inclusion criteria were: male or female; ≥ 70 years old; emergency admission with primary fracture of the hip and able to gain written informed consent from the patient or personal consultee.
Initial recruitment (n = 100) was solely patients able to consent for themselves. Following study amendment, the final 241 participants included both those unable to give their own consent and those who could provide consent. Exclusion criteria were: terminal illness, pathological fracture, in-hospital fractures, and patients enrolled in a medicinal product interventional trial.
Study items of interest and criteria definitions were initially identified from both the original POMS 5 7 and C-POMS 12 . A literature search, to identify any additional commonly observed complications after hip fracture surgery was performed. A draft list of potential additional and amended items was circulated to six orthogeriatricians for their views, comments and addition of items they considered important from clinical experience.
Three steps were taken to identify redundant items, a) a pairwise correlation matrix of all items on the data collection form was performed, b) item prevalence within the study population and c) all new identified items were sent to a team of orthogeriatricians for comments on item clinical severity, importance and the likelihood of currently being captured by the POMS [7]. 6 Items with a correlation of > 0.8 were excluded. Using feedback and comments from the specialty team, items for inclusion had to meet three of the following criteria: The identified morbidity item prevalence must be ≥ 5% (ii) Whether the morbidity item identified is likely to be captured by the original POMS or otherwise. The likelihood was measured using a 5-point Likert scale with threshold for inclusion set as median < 3 (i.e. unlikely to be captured) (iii) The likelihood that the patient remains in acute care due to this morbidity item. Society of Anaesthesiology physical assessment (ASA) 18 19 . Mobility was measured using a modified Cumulated Ambulation Score (CAS) 20  for pain related to surgery, excluding codeine phosphate and dihydrocodeine. Patients with higher perioperative risk scores (NHFS > 4 and those with ASA grades III-V) developed more HF-POMS defined morbidities (appendix 2) and had increased subsequent LOS. Any increase in NHFS and ASA predicted an increase in LOS by 2.0 days (95% CI 0.5 -3.5, p = 0.008) and 2.5 days (95% CI 0.6 -4.5, p = 0.012) respectively.

Other reasons identified for staying in hospitals
Social reasons (appendix 3) are additional to non-medical reasons for why some of the patients remained in hospital.

Discussion
Our study results support the development of a hip fracture specific POMS. The pattern of morbidity following hip fracture is clinically different to the original POMS: HF-POMS demonstrates the expected relationships with time, known predictors of risk and presence of HF-POMS is predictive of prolonged length of stay. In common with previous work, and patient and clinical experience, a significant number of patients remain in acute hospital care following fracture without overt medical reason.
The 2.4% in-hospital mortality rate is in line with previous studies 25 and the average LOS of 16.9 days similar to the national average 1 . The overall studied population was biased slightly towards a fitter group, due to the initial recruitment of participants with capacity only. There are differences in local practice around acute hospital discharge between the three units, reflected in the differing lengths of stay.
Higher risk patients have greater morbidity 19 and in turn, morbidity is predictive of subsequent length of stay 26 . The HF-POMS has face validity as it has been developed solely from emergency hip fracture patients and morbidity domains identified are commonly known hip fracture complications 27 known to associate with LOS [28][29][30] .
Studies that validated the original POMS 3 6-8 in various surgeries have shown that individual POMS domains can predict LOS. The domain associated with greatest increase in length of stay (endocrine) was relatively infrequent (6%) and so is probably less clinically relevant. In contrast, impaired mobility has the greatest overall impactwith a large proportion of patients, and a moderate increase in LOS. Grocott  in an effort to reduce the risk. 14 A few study limitations were observed in this study. Our age inclusion criterion was 70 years and above which could limit generalisation of the results to patient below this age; this is a small proportion of patients with hip fracture. Seventy-one percent of participants from one of the study sites (UHL) had their NHFS missing as the data is not routinely collected. Lowmoderate internal consistency was observed among the morbidity domains suggesting that the tool might not be suitable to be used as a summary score reflecting the views of Grocott 7 .
The increased serum creatinine levels as defined by the HF-POMS were observed in a very low number of participants (only 1%). This could have been due to the limited access to biochemistry results available for the study team. Furthermore, such blood tests are not a daily routine. Additionally, although minor degrees of acute kidney injury are a risk factor for poorer outcome, it is asymptomatic and seem unlikely to be a cause for staying in acute care per se 34 .
Data for HF-POMS were collected from three NHS sites and inter-rater reliability was good reflecting its potential for nationwide application. All research sites were acute care trusts and morbidity was measured during hospital stay. As a standard discharge guide framework, HF-POMS could be used to categorise patients at different levels within the discharge pathway improving early discharge particularly with the current bed pressures in the NHS.
Moreover, this might determine discharge destination, whether the patient will be transferred from acute care to a community hospital, intermediate care or residential home for further rehabilitation or they will be discharged back to their own home.
The HF-POMS does not address non-medical aspects affecting prolonged hospital stay.
Several non-morbidity factors have been identified keeping the patients in hospital 7 12 . In this study we observed 8% remained in hospital for social reasons only on day 15.
In the light of the validation provided by this study we believe that HF-POMS has potential value for local quality improvement and audit, commissioning and research. At an immediate 15 practical level, it may assist members of the hip fracture care pathway in discussing and making discharge plans. We have described the variation in morbidity factors at different stages post-surgery. High morbidity scores at any post-operative point are associated with a longer length of stay.