John Kellett

An improved medical admissions risk system using multivariable fractional polynomial logistic regression modelling

AIM To develop and validate an in-hospital mortality risk prediction tool for unselected acutely ill medical patients using routinely collected physiological and laboratory data. DESIGN Analysis of all emergency medical patients admitted to St Jamess Hospital (SJH), Dublin, between 1 January 2002 and 31 December 2007. Validation using a dataset of acute medical admissions from Nenagh Hospital 2000-04. METHODS Using routinely collected vital signs and laboratory findings, a composite 5-day in-hospital mortality risk score, designated medical admissions risk system (MARS), was developed using an iterative approach involving logistic regression and multivariable fractional polynomials. Results are presented as area under receiver operating characteristics curves (AUROC) as well as Hosmer and Lemeshow goodness-of-fit statistics. RESULTS A total of 10 712 and 3597 unique patients were admitted to SJH and Nenagh Hospital, respectively. The final score included nine variables [age, heart rate, mean arterial pressure, respiratory rate, temperature, urea, potassium (K), haematocrit and white cell count]. The AUROC for 5-day in-hospital mortality was 0.93 [95% confidence interval (CI) 0.92-0.94] for the SJH cohort (Hosmer and Lemeshow test, P = 0.32) and 0.92 (95% CI 0.90-0.94) for the external Nenagh hospital validation cohort (Hosmer and Lemeshow test, P = 0.28). CONCLUSION In-hospital mortality estimation using only routinely collected emergency department admission data is possible in unselected acute medical patients using the MARS system. Such a score applied to acute medical patients at the time of admission, could assist senior clinical decision makers in promptly and accurately focusing limited clinical resources. Further studies validating the impact of this model on clinical outcomes are warranted.

Derivation and validation of a score based on Hypotension, Oxygen saturation, low Temperature, ECG changes and Loss of independence (HOTEL) that predicts early mortality between 15 min and 24 h after admission to an acute medical unit

BACKGROUND Predictive scores such as APACHE II have been used to assess patients in intensive care units, but few scores have been used to assess acutely ill general medical patients. DESIGN Examination of the ability of clinical variables documented at the time of admission to predict early mortality between 15 min and 24 h after admission. SETTING An Irish rural hospital. SUBJECTS 10,290 consecutive patients admitted as acute medical emergencies, divided into a derivation cohort of 6947 patients and a validation cohort of 3343 patients. RESULTS 40 patients of the derivation cohort (0.6%) died within 24h of hospital admission. Multivariate analysis revealed 11 independent predictors of early death from which a simplified model with minimal loss of predictive ability was derived. Since this model contained only the five variables of Hypotension (systolic blood pressure<100 mm Hg), low Oxygen saturation (<90%), low Temperature (<35 degrees C, abnormal ECG and Loss of independence (unable to stand unaided) it was named the HOTEL score (one point for each variable). There were no differences in the early mortality predicted by this score between the derivation and validation cohorts-the area under the receiver operator characteristic curves for the derivation and validation cohorts were 86.5% and 85.4%, respectively. None of the patients with a score of zero died within 15 min and 24 h and a score of one had an early mortality of 0.3% in both cohorts. A score of two had an early mortality of 0.9% in the derivation cohort and 1.7% in the validation cohort, while a score of three or greater had an early mortality of 10.2% in the derivation and 5.6% the validation cohort. CONCLUSIONS The HOTEL score quickly identifies patients at a low and high risk of death between 15 min and 24 h after admission, thus enabling prompt triage and placement within a health care facility.

Impedance cardiography — A rapid and cost-effective screening tool for cardiac disease

Impedance cardiography (ICG) charts the rises and falls of thoracic impedance as the fluid content of the chest changes with each heartbeat. Breathing, arrhythmia, movements and posture interfere with the ICG. Modern pattern recognition software can now produce a composite signal-averaged ICG that considerably simplifies interpretation. The first derivative velocity waveform shows a smooth S wave that corresponds with systole, while the second derivative acceleration waveform (dZ/dt) contains several reference points that outline the A, S and O waves. Normally, the A wave follows atrial contraction and occurs in late diastole. It can, therefore, be abnormal in both atrial and ventricular arrhythmias and is abnormally increased when there is diastolic dysfunction. The S wave reflects ventricular contractility and is deformed by ventricular dyssynchrony. The O wave is associated with mitral valve opening and is abnormally enlarged in heart failure. These different patterns of ICG waveform are relatively easy to recognise and can be cost-effectively and quickly obtained to reliably distinguish between normal and abnormal cardiac function.

Changes and their prognostic implications in the abbreviated Vitalpac™ early warning score (ViEWS) after admission to hospital of 18,853 acutely ill medical patients

BACKGROUND The best performing early warning score is Vitalpac™ Early Warning Score (ViEWS). However, it is not known how often, to what extent and over what time frame any early warning scores change, and what the implications of these changes are. SETTING Thunder Bay Regional Health Sciences Center, Ontario, Canada. METHODS The changes in the first three complete sets of the six variables required to retrospectively calculate the abbreviated version of ViEWS (that did not include mental status) after admission to hospital of 18,853 acutely ill medical patients, and their relationship to subsequent in-hospital mortality were examined. RESULTS In the 10.4 SD 20.1 (median 5.0) hours between admission and the second recording the score changed in only 5.9% of patients and these changes were of no prognostic value. By the time of the third recording 34.9 SD 21.7 (median 30.0) hours after admission a change in score was clearly associated with a corresponding change in in-hospital mortality (e.g. for patients with an initial score of 5 an increase between the first and third recording of ≥4 points was associated with an increased mortality (OR 6.5 95% CI 2.3-15.9, p<0.00001), whereas a reduction of ≤-4 points was associated with a reduced mortality (OR 0.4 95% CI 0.2-0.9, p 0.03)). CONCLUSION After a median interval of 30 h both the initial abbreviated ViEWS recording and subsequent changes in it both predict clinical outcome. It remains to be determined what interventions during this time frame will improve patient outcomes.

Validation of the VitalPAC™ Early Warning Score (ViEWS) in acutely ill medical patients attending a resource-poor hospital in sub-Saharan Africa.

BACKGROUND The VitalPAC™ Early Warning Score (ViEWS) has an area under the receiver operator characteristic curve (AUROC) for death of acute unselected medical patients within 24h of 88% and the UK National Early Warning Scores is based on it. The scores discrimination has been validated on patients in the developed world, but nothing is known of its performance in resource-poor hospitals. METHODS ViEWS was validated in 844 acutely ill medical patients admitted to Kitovu Hospital, Masaka, Uganda. RESULTS The AUROC for death within 24h of admission was 88.6% (95% CI 82.5-94.7%). The inability to walk without help was found to be an additional independent predictor of in-hospital mortality, and ViEWS modified to include it had an AUROC for death within 24h of 91.9% (95% CI 86.5-97.2%). CONCLUSION The discrimination of ViEWS in a resource poor sub-Saharan Africa hospital is the same as in the developed world. Inability to walk without help was found to be an additional independent predictor of mortality.

Early Warning Scores Generated in Developed Healthcare Settings Are Not Sufficient at Predicting Early Mortality in Blantyre, Malawi: A Prospective Cohort Study

Aim Early warning scores (EWS) are widely used in well-resourced healthcare settings to identify patients at risk of mortality. The Modified Early Warning Score (MEWS) is a well-known EWS used comprehensively in the United Kingdom. The HOTEL score (Hypotension, Oxygen saturation, Temperature, ECG abnormality, Loss of independence) was developed and tested in a European cohort; however, its validity is unknown in resource limited settings. This study compared the performance of both scores and suggested modifications to enhance accuracy. Methods A prospective cohort study of adults (≥18 yrs) admitted to medical wards at a Malawian hospital. Primary outcome was mortality within three days. Performance of MEWS and HOTEL were assessed using ROC analysis. Logistic regression analysis identified important predictors of mortality and from this a new score was defined. Results Three-hundred-and-two patients were included. Fifty-one (16.9%) died within three days of admission. With a cut-point ≥2, the HOTEL score had sensitivity 70.6% (95% CI: 56.2 to 82.5) and specificity 59.4% (95% CI: 53.0 to 65.5), and was superior to MEWS (cut-point ≥5); sensitivity: 58.8% (95% CI: 44.2 to 72.4), specificity: 56.2% (95% CI: 49.8 to 62.4). The new score, dubbed TOTAL (Tachypnoea, Oxygen saturation, Temperature, Alert, Loss of independence), showed slight improvement with a cut-point ≥2; sensitivity 76.5% (95% CI: 62.5 to 87.2) and specificity 67.3% (95% CI: 61.1 to 73.1). Conclusion Using an EWS generated in developed healthcare systems in resource limited settings results in loss of sensitivity and specificity. A score based on predictors of mortality specific to the Malawian population showed enhanced accuracy but not enough to warrant clinical use. Despite an assumption of common physiological responses, disease and population differences seem to strongly determine the performance of EWS. Local validation and impact assessment of these scores should precede their adoption in resource limited settings.

Readmissions of medical patients: an external validation of two existing prediction scores

BACKGROUND Hospital readmissions are increasingly used as a quality indicator with a belief that they are a marker of poor care and have led to financial penalties in UK and USA. Risk scoring systems, such as LACE and HOSPITAL, have been proposed as tools for identifying patients at high risk of readmission but have not been validated in international populations. AIM To perform an external independent validation of the HOSPITAL and LACE scores. DESIGN An unplanned secondary cohort study. METHODS Patients admitted to the medical admission unit at the Hospital of South West Jutland (10/2008-2/2009; 2/2010-5/2010) and the Odense University Hospital (6/2009-8/2011) were analysed. Validation of the scores using 30 day readmissions as the endpoint was performed. RESULTS A total of 19 277 patients fulfilled the inclusion criteria. Median age was 67 (range 18-107) years and 8977 (46.6%) were female. The LACE score had a discriminatory power of 0.648 with poor calibration and the HOSPITAL score had a discriminatory power of 0.661 with poor calibration. The HOSPITAL score was significantly better than the LACE score for identifying patients at risk of 30 day readmission (P < 0.001). The discriminatory power of both scores decreased with increasing age. CONCLUSION Readmissions are a complex phenomenon with not only medical conditions contributing but also system, cultural and environmental factors exerting a significant influence. It is possible that the heterogeneity of the population and health care systems may prohibit the creation of a simple prediction tool that can be used internationally.

Comparison of the heart and breathing rate of acutely ill medical patients recorded by nursing staff with those measured over 5 min by a piezoelectric belt and ECG monitor at the time of admission to hospital

BACKGROUND Heart and breathing rates are predictors of disease severity and of a poor outcome. However, few reports have compared their machine measurements with traditional manual methods. SETTING A small rural Irish hospital. METHODS The heart and breathing rates of 377 acutely ill medical patients (mean age 68.3 SD 16.8 years) recorded by nursing staff at the time of admission to hospital was compared with those measured over 5 min by a piezoelectric belt and ECG monitor (the BT16 acquisition system). RESULTS The mean breathing rate measured by the nursing staff (20.9 SD 4.8 breaths per min) and that measured by the BT16 piezoelectric belt (19.9 SD 4.5 breaths per min) were significantly different (p 0.004), as were the nurse and BT16 measured heart rates (85.4 SD 21.3 vs. 81.2 SD 18.7, p 0.004), and the correlation coefficient between the two methods of breathing and heart rate measurement were low. Nurse measured breathing rate measurements were clustered around rates of 18, 20 and 22 breaths per min. Unlike those obtained by nurses, BT16 measured heart and breathing rates were shown by logistic regression to be independent predictors of in-hospital mortality. CONCLUSION There is a poor correlation between breathing and heart rates measured by traditional methods and those obtained by the BT16 device. BT16 derived breathing and heart rates, but not those measured manually, were independent predictors of in-hospital mortality.

Collaborative Audit of Risk Evaluation in Medical Emergency Treatment (CARE-MET I) — An international pilot

BACKGROUND The absence of an accepted model for risk-adjustment of acute medical admissions leads to suboptimal clinical triage and serves as a disincentive to compare outcomes in different hospitals. The Simple Clinical Score (SCS) is a model based on 16 clinical parameters affecting hospital mortality. METHODS We undertook a feasibility pilot in 21 hospitals in Europe and New Zealand each collecting data for 12 or more consecutive medical emergency admissions. Data from 281 patients was analysed. RESULTS Severity of illness as estimated by SCS was related to risk of admission to the Intensive Care Unit (p<0.001) but not to the Coronary Care Unit. Mortality increased from 0% in the Very Low Risk group to 22% in the Very High Risk Group (p<0.0001). Very low scores were associated with earlier discharge as opposed to very high scores (mean length of stay of 2.4 days vs 5.6 days, p<0.001). There were differences in the pattern of discharges in different hospitals with comparable SCS data. Clinicians reported no significant problems with the collection of data for the score in a number of different health care settings. CONCLUSION The SCS appears to be a feasible tool to assist clinical triage of medical emergency admissions. The ability to view the profile of the SCS for different clinical centres opens up the possibility of accurate comparison of outcomes across clinical centres without distortion by different regional standards of health care. This pilot study demonstrates that the adoption of the SCS is practical across an international range of hospitals.

Who will be sicker in the morning? Changes in the Simple Clinical Score the day after admission and the subsequent outcomes of acutely ill unselected medical patients

BACKGROUND All doctors are haunted by the possibility that a patient they reassured yesterday will return seriously ill tomorrow. We examined changes in the Simple Clinical Score (SCS) the day after admission, factors that might influence these changes and the relationship of these changes to subsequent clinical outcome. METHOD The SCS was recorded in 1165 patients on admission and again the following day (i.e. 25.0±15.8 h later). The abilities of 51 variables that might predict changes in the SCS were examined. RESULTS The day after admission 16.1% of patients had been discharged home, 31.4% had decreased their SCS by 2.4±1.6 points, 38.6% had an unchanged SCS, 12.0% had increased their SCS by 2.1±1.7 points and 1.2% had died. Patients with an increased SCS had higher in-hospital mortality (10% vs. 1.1%, OR 10.1, p<.001) and a longer length of stay (9.4±9.6 vs. 5.6±7.0 days, p<.001). There was no consistent association between the SCS recorded at admission and SCS increase. Only nursing home residence, heart failure and a Medical Admission Risk System laboratory data score>0.09 were found to be independent predictors of SCS increase. CONCLUSION The SCS of 12% of patients increases the day after admission to hospital, which is associated with a ten-fold increase of in-hospital mortality. Low SCS risk patients are just as likely to have a SCS increase as high risk patients.