Weekly Hospital Workforce Data: A Data Visualisation Exercise Yang - - PDF document

Weekly Hospital Workforce Data: A Data Visualisation Exercise

Yang XIEa, Sankalp KHANNAa, Norm GOODa and Justin BOYLEa

a The CSIRO Australian e-Health Research Centre, Brisbane, Australia

• Abstract. Quantifying the health workforce in terms of overall staff numbers and

their ratio to patients under their care can strengthen analytical studies designed to inform policy regarding how hospital services are delivered. Information about staffing is traditionally obtained via location-specific audits or self-reported information gleaned from surveys which hold potential biases around time- dependence and recall. In contrast, work presented in this paper describes the derivation of useful workforce metrics from routine hospital financial and clinical information systems that overcome these biases. Staffing data is aggregated, visualised and linked to patient demand to gain insight into spatial and temporal variations in hospital staffing and workload. Overall, hospital staff resourcing varies noticeably across a week, with staff numbers and staff-to-patient ratios dropping to low levels at night and across a weekend. Exploration of staff-to-staff ratios allows further insight into staff dynamics across a week and the variation of supervision level.

• Keywords. Staff visualisation, patient ratios, health workforce, hospital

performance

Introduction Effective staffing of hospitals directly impacts their financial, safety and quality, and bed access performance. Fiscal impacts are obvious, as staffing costs are among the largest categories in hospitals’ budgets, where nursing staff alone have been estimated to account for 25% or more of annual operating expenses and as much as 40% of direct care costs [1]. Workforce issues have also been noted to be an issue with hospital crowding [2, 3]. Inadequate staffing has been described as one of the most obvious factors related to hospital overcrowding [4]; a major limiting factor for staffed bed availability [5]; and their being integral to a fully functioning hospital system. Shortage

• f staff has been stated as causing increased workloads, precipitating high turnover, and

a resultant disproportionate level of inexperienced replacement personnel. Redeployed staff may fill the numbers but are working in unfamiliar terrain; with effects being manifested in productivity when overstretched clinicians attempt to make up the difference, with a threat to patient care [3]. Excessive or inappropriate workloads can also result in loss of hospital staff. An Australian study [6] calculated the average ED nurse to patient ratio of 1 to 15 on a morning shift, and found that staff modify practice in order to cope with such demand. While this adaptation ensures survival in the short term, the long-term implications are burnout, followed by leaving, with resultant fiscal and competency losses for the system.

There has been evidence reported of an association between lower staff workloads and better patient outcomes, including lower hospital mortality [7-11]. Research has shown that a higher number of registered nurses relative to the number of patients has a positive impact on patient outcomes including decreased lengths of stay in hospital. Evidence also indicates that appropriate staffing numbers benefits the workforce by reducing work-related injuries, absenteeism and turnover, and by increasing job satisfaction [12]. If hospital staff are to provide timely and high-quality care [13], then care and attention need to be paid to their numbers and ratios with respect to patient demand. Many analytical studies in the field of health informatics aiming to turn data and metrics into policy would benefit from inclusion of workforce data. For example, in a controversial study aimed at generating evidence around differential outcomes for patients admitted to hospital on weekends [14], a highly emotive debate has exploded in published responses to the article claiming that the underlying statistical models do not incorporate information related to how hospitals are staffed. To the best of our knowledge, studies that include workforce metrics have involved location-specific audits or relied on large mail surveys and self-reported workload assessment, rather than deriving this from hospital administration systems. For example, Aiken et al [1, 15-16] describe their calculation of workforce data from surveys of staff by dividing the average number of patients reported by staff in a particular hospital unit on their last shift by the average number of staff on the unit for that same shift. These methods are costly and can be biased based on the time they are carried out as well as by recall issues and motives held by survey participants. Our team have been working on generating evidence in relation to the delivery of acute health services and wish to include workforce data in the analyses to present policy makers with the best possible evidence. In most hospitals, information related to staffing is maintained within financial administration systems. However the form of workforce data is unwieldy, being mainly designed for payroll purposes. This paper describes the reshaping, aggregation and visualisation of workforce data and its linking with patient demand to create several useful metrics that have application to informing policy on the way healthcare should be delivered.

• 1. Methods

This study used extracts of hospital admission data and payroll records covering the 28 largest public reporting hospitals in Queensland, Australia from April 2013-December

• 2015. The 28 facilities were categorised into 6 hospital peer groups in accordance with

AIHW public hospital peer groupings. Admissions data covered approximately 3 million admission records from 1.1 million unique patients, and payroll data comprised approximately 34 million records capturing variables representing start and end time of shift, facility name, the role and hierarchical level of staff as well as worked hours and minutes. The raw workforce payroll records reflecting start and end times of individual staff were converted into counts of staff present at various hierarchical and operational areas at an hourly time resolution throughout the study period. The process used for calculating ratios was designed to be replicable and automated for generalising to a range of analytical studies. First, every payroll record was segmented into hourly slots. For example, a payroll record for a particular staff member spanning from 2011-03-17

8:15:00 to 2011-03-17 10:00:00 would be segmented into two hourly slots (2011-03-17 8:00:00 - 2011-03-17 9:00:00) and (2011-03-17 9:00:00 - 2011-03-17 10:00:00). In these two hourly slots, the individual staff member worked 0.75 hour and 1 hour respectively, thereby contributing to 0.75 staff and 1 staff in these two hourly slots. After segmenting all payroll records, aggregation by facility, position and hourly slot start timestamp and end timestamp was applied. A robust derivation of the numbers of every staff role in each hour in each facility across the study period was calculated in this way. Staffing data was categorised by the project’s Advisory Group to focus on clinical staff in attendance during a shift i.e., the available care team, whose presence was deemed to make a difference in care and also seniority – i.e., nursing and doctor classifications were separated into junior/intermediate/senior, as per Table 1.

Table 1. Role classification description. Classification Description A&O Clinical Administrative and Operational Staff with clinical responsibilities HP Clinical Heath Practitioners with clinical responsibilities Junior Nurses Trainee, Assistant, Student and Enrolled Nurses Intermediate Nurses Registered, Clinical & Consultant Nurses, Clinical Educators & Nurse Practitioners Senior Nurses Nurse Directors and Assistant, Nursing and Executive Directors of Nursing Junior Doctors Resident Medical Officers Intermediate Doctors Medical Registrars Senior Doctors Medical Senior Officers, Staff Specialists, Visiting Medical Offers and Specialists, and Superintendents Non-Clinical Senior Admin and Operational Staff and Health Practitioners, Professional and Technical Staff and Senior Health Executives

Each row in the reshaped payroll data (i.e., hourly staff data) represented an hourly slot in a particular facility, and contained counts of staff of a particular role, working in that facility during that hourly slot. Hospital admission records were reshaped using a similar method as reshaping payroll data. Each row in the reshaped admission data (i.e., hourly patient data) represented an hourly slot in a particular facility and contained counts of inpatients staying in that facility during that hourly slot. After the reshaping, hourly admission data and hourly workforce data were then joined by facility name and hourly slot, thereby deriving an estimate of both the staffing levels (number of nurses, doctors, etc.) and number of patients at a particular facility in each hour. The resulting reshaped datasets reflected continuous counts of data in every hour across the study period, which were then aggregated to a weekly level by taking means. Therefore, estimates of average number of staff numbers working in every hour in a week in different hospital peer groups were derived.

Additional workforce metrics were generated based on variables listed in Table 1. The dataset covered facilities of different sizes, and in order to eliminate the effect of hospital size, two types of staff ratio features were calculated: staff-staff ratio and staff- patient ratio. Staff-patient ratio reflected the level of staff-to-patient resources. For example, if there were 4 junior doctors and 20 patients in hospital A in an hourly slot, then the junior-doctor-to-patient ratio (Junior.Doctors.Pat.Ratio) in that hour in hospital A would be 4/20 = 0.2. This staff-patient ratio can be calculated for all 9 staff roles listed in Table 1, plus an additional metric covering all staff-to-patient ratio. In this way, 10 staff-patient ratio features were generated. Staff-staff ratio reflected the level of one staff group relative to another. For A&O clinical, HP clinical and non-clinical staff, this staff-staff ratio measures the proportion

• f staff with such roles among all staff. For example, if there were 10 A&O clinical

staff and 100 all-kind staff working in hospital A in an hourly slot, then the A&O- clinical-staff-to-all-staff ratio (A.O.Clinical.Ratio) in that hour in hospital A would be 10/100 = 0.1. For nurses and doctors, their staff-staff ratio were derived in a different

• way. There were 3 hierarchical levels defined for nurses and doctors (i.e., junior,

intermediate, and senior), and their staff-staff ratio was designed to capture supervision

• level. The staff ratio for a nurse or doctor at a particular hierarchical level was derived

by the number of that level nurses or doctors divided by the total number of all level nurses or doctors. For example, if there were 10 junior doctors and 40 doctors of all levels working in hospital A in an hourly slot, then the junior doctor ratio (Junior.Doctors.Ratio) in that hour in hospital A would be 10/40 = 0.25. In this way, 9 staff-staff ratio features were generated in total.

• 2. Results

Following the merging of financial and patient information, several valuable workforce metrics were generated) for every hour in a week, by hospital peer group. Figure 1 shows representative figures of these workforce metrics. a) Average number of doctors (junior, intermediate, and senior) b) Average number of nurses (junior, intermediate, and senior) c) Average number of the other staff roles (AO clinical, HP clinical and Non- clinical) d) Doctor/patient ratio (junior, intermediate and senior) e) Nurse/patient ratio (junior, intermediate and senior) f) Other staff/patient ratios (AO clinical, HP clinical and Non-clinical) g) Doctors ratio (junior, intermediate and senior) h) Nurses ratio (junior, intermediate and senior) i) Other staff ratios (AO clinical, HP clinical and Non-clinical)

• 3. Discussion

Visualisation of workforce data as presented in this paper illustrates the patient-driven workload associated with acute healthcare delivery. From Figure 1, it can be seen that,

• verall, all staff (i.e. doctors, nurses and other staff) experienced a sharp decline in

numbers across a weekend. Differences between hospital peer groups are not shown in

Figure 1, but the larger hospitals experienced the most dramatic decline of doctors on a

• weekend. During peak periods, the larger hospitals also had similar numbers of senior

and intermediate doctors, while in smaller hospitals, the senior doctors dominated in

• numbers. Most doctors at night were intermediate doctors.

Our visualisations have also determined that across all hospitals, the hierarchical proportions of nurses were similar across hospital peer groups, and the proportion of intermediate nurses was usually the highest, followed by junior and senior nurses. Senior nurses are rarely seen at night. Other staff roles also vary across a week dropping to a very low level at night and during a weekend. Staff-patient ratios exhibit a similar pattern with a decline at night and in weekends. Children’s hospitals were

• bserved to have the highest doctor-to-patient and nurse-to-patient ratios. In contrast,

public acute group B hospitals had the lowest values of these two staff-patient ratios.

a b c d e f

Figure 1. Example plots showing the visualisation for the workforce metrics. The data represent mean values across the study period for every hour within a week from Monday 12AM-1AM (Hour 1) to Sunday 11PM- 12AM (Hour 168).

The staff:staff ratios allows further insight into staff dynamics across a week. For example, we observe that the senior doctor ratio drops at night and on the weekend, when instead the intermediate doctor ratio dominates and becomes the major working force after hours. The junior doctor ratio remained stable across weekdays and weekends except in small hospitals where the junior doctor ratio declined over a weekend and the senior doctor ratio remained dominant in both normal hours and after

• hours. Public acute group B hospitals also have different hierarchical proportions of

doctors, with junior doctors occupying a large proportion. Intermediate nurses were the dominant group of nurses in all peer groups and across all times of the week. The senior nurse ratio was observed to drop after hours which could reflect less supervision. The usefulness of this data aggregation has been extended by clustering the staff ratios (using K-means) into two groups (high and low values). This clustering almost perfectly separated normal hours and after hours (i.e., nights and weekends), highlighting the applicability of these metrics for subsequent statistical modelling (for example, their inclusion in models to assess differences in patient outcomes by admission time).

• 4. Conclusion

Reshaping hospital payroll data, aggregation to an hourly time interval and linking to patient admission data has enabled the creation of useful metrics to model hospital service delivery. Derivation of these metrics in this way overcomes the cost and biases associated with audit and survey approaches, and as a result, improved insight into staff dynamics and workforce variations has been facilitated. The methodology presented in this paper can be automated for use at scale to help investigate how staffing level may

i g h

affect health outcomes for patients in conjunction with other important demographic and clinical variables in the future. References

[1] Aiken L, Cimiotti J, Sloane D, et al. The Effects of Nurse Staffing and Nurse Education on Patient Deaths in Hospitals With Different Nurse Work Environments. Med Care. 2011 Dec; 49(12): 1047– 1053. [2] Emergency Department Crowding: High Impact Solutions, Emergency Medicine Practice Committee, American College of Emergency Physicians, May 2016; Available online [AccessedMay 2017]:: https://www.acep.org/Legislation-and-Advocacy/Practice-Management-Issues/Boarding/Crowding/ Emergency-Department-Crowding---High-Impact-Solutions/. [3] Derlet R, Richards J. Overcrowding in the nation's emergency departments: complex causes and disturbing effects. Ann Emerg Med. 2000 Jan;35(1):63-8. [4] Howard K. Overcrowding: not just an emergency department issue. J Emerg Nurs. 2005 Jun;31(3):227- 8. [5] Asplin B, Magid D, Rhodes K, et al. A conceptual model of emergency department crowding. Ann Emerg Med. 2003 Aug;42(2):173-80. [6] Lyneham J, Cloughessy L, Martin V. Workloads in Australian emergency departments a descriptive

• study. Int Emerg Nurs. 2008 Jul;16(3):200-6.

[7] Kane R, Shamliyan T, Mueller C, et al. The association of registered nurse staffing levels and patient

• utcomes: systematic review and meta-analysis. Med Care. 2007;45:1195–1204.

[8] Aiken L, Clarke S, Sloane D, et al. Effects of hospital care environments on patient mortality and nurse

• utcomes. J Nurs Adm. 2008;38:220–226.

[9] Van den Heede K, Lesaffre E, Diya L, et al. The relationship between inpatient cardiac surgery mortality and nurse numbers and educational level: analysis of administrative data. Int J Nurs Stud. 2009;46:796–803. [10] Needleman J, Buerhaus P, Pankratz S, et al. Nurse staffing and inpatient hospital mortality. N Engl J

• Med. 2011;364:1037–1045.

[11] Blegen M, Goode C, Spetz J, et al. Nurse staffing effects on patient outcomes: safety-net and non- safety-net hospitals. Med Care. 2011;49:406–414. [12] Queensland Health. Office of the Chief Nursing and Midwifery Officer. Nurse-to-patient ratios. May 2016, available online: https://www.health.qld.gov.au/__data/assets/pdf_file/0027/357453/ratiosqa.pdf [accessed May 2017]. [13] Australasian College for Emergency Medicine. Policy on The Australasian Triage Scale, Policy Document P06 v4, July 2013, available online: https://acem.org.au/getattachment/693998d7-94be-4ca7- a0e7-3d74cc9b733f/Policy-on-the-Australasian-Triage-Scale.aspx [accessed May 2017]. [14] Freemantle N, Ray D, McNulty D, et al. Increased mortality associated with weekend hospital admission: a case for expanded seven day services? BMJ. 2015 Sep 5;351:h4596. [15] McHugh M, Rochman M, Sloane D, et al. Better Nurse Staffing and Nurse Work Environments Associated With Increased Survival of In-Hospital Cardiac Arrest Patients, Med Care. 2016 Jan; 54(1): 74–80. [16] Cho E, Sloane D, Kim EY, et al. Effects of nurse staffing, work environments, and education on patient mortality: An observational study. Int J Nurs Stud. 2015 Feb; 52(2): 535–542.