STUDY OF CORRELATION OF HOSPITAL NURSE STAFFING AND OUTCOMES Tentative Study Approach March 25, 2014
Overview Reality Check Data availability on staffing and outcomes Data about confounding factors Chronology of staffing and outcome data Tentative Staged Approach Short Term: Variation in staffing Short Term: Association between available staffing and outcome data Mid-term: Analysis of additional staffing data to overcome shortcomings Potentially longer-term: Analysis w/confounding variables 2
Data Availability: Staffing Data Staffing plan data available in Spring 2014 (for CY/FY 2014) Actual staffing data in June 2014 for the first quarter Data will be at the unit level (categories from the Labor Management Institute) Data will not be: Specific to shifts, including potential high-patient load periods Distinguish between various levels of qualifications/experience for nurses Break out other assistive personnel into the granularity favored in workgroup discussions 3
Data Availability: Confounding Factors Case-mix adjusters are built-in to standard groupers, such as APR-DRGs Patient complexity can be ascertained through analysis of the diagnostic codes Patient demographics may further help to make valid comparisons (age, gender, zip code, etc.) Limits: True acuity systems do not exist uniformly and they vary in their robustness Data on factors such as unit activity, culture/environment, nurse autonomy are not readily available and challenging to collect across the entire hospital industry 4
Data Availability: Outcome Data About a dozen outcome measures exists that are nursesensitive Some concordance (not perfect) across various measure owners: infection, falls, pressure ulcers Many measures are claims-based Data are not: Specific to units or shifts, but measure an overall hospital s performance Fully reflective of the range of nurse staffing performance Real-time because many are based on claims data, they come with a lag Characterized by large number of observations short periods of data will now allow identification of associations 5
Chronology of Data Data on patient outcomes, hospital claims and nurse staffing are poorly aligned over time Association between variables are only meaningful if relationships are sticky over time -- 2012 -- -- 2013 -- -- 2014 -- Jan Jan Claims Data Dec Oucome Data Dec Jan Mar Staffing Data 6
TENTATIVE STUDY APPROACH 7
MDH Nurse Staffing Levels and Patient Outcomes - Study Timeline 8
Tentative Study Approach Overview Short term (April June 2014) Part 1: What can available data tell us about hospital nurse staffing variation in Minnesota across peer hospitals? Part 2: To what extent can variation in this preliminary staffing data tell us about the relationship between staffing and patient outcomes? Medium term (July October 2014) What additional analysis and data collection is necessary to strengthen the preliminary analysis and overcome some of the limitations associated with available data? Potential long term? (2015 and beyond) How to obtain data to address confounders? 9
SHORT-TERM APPROACH April June 2014 10
Part 1: Preliminary Description of Nurse Staffing Variation in Minnesota MHA Nurse Staffing Plans (April 2014) Inpatient nurse staff only Non-managerial staff direct patient care No skill mix information Staffing by unit type: ICU and non-icu Hospital Annual Report (HAR) vintage FY 2012 and historical All hospital nurse staff (inpatient and outpatient) Skill mix: RN, LPN, Nursing Assistant/Aide Revenue and utilization data 11
Part 2: Preliminary Analysis of the Relationship Between Staffing and Patient Outcomes Data HAR (most recent FY 2012) Patient length of stay (inpatient or adjusted) Hospital Adverse Health Events (most recent FY 2013) Patient fall with serious injury or death All patient falls? Stage 3 or 4 pressure ulcers (serious bedsores) Hospital Discharge Data (most recent CY 2012) for National Quality Forum Nurse Sensitive Core Measures Decubitus/pressure ulcer Failure to rescue Infections due to medical care (ICU only) 12
Nurse Sensitive Patient Outcome Measures Select Measure Name Death in low mortality DRG Decubitus/ pressure ulcer AHRQ Patient Safety Indicators NQF Core Measures ANA National Database of NQI Potential Data Sources (Vintage) X HDD (2012) X X X HDD (2012)/ HAE (2013) Failure to rescue X X X HDD (2012) Postoperative PE or DVT (PSI 12) Infection due to medical care Patient falls prevalence X HDD (2012) X Only ICU X HDD (2012) X HAE (2013) Patient falls with injury X HAE (2013) Restraint prevalence Source: Adapted from Savitz et al., Quality indicators sensitive to nurse staffing in acute care settings. In: Hendrickson K, Battles JB, Marks ES, et al, eds. Advances in Patient Safety: From Research to Implementation. Rockville, MD: Agency for Healthcare Research and Quality; 2005:375 385 Note: HDD is Hospital Administrative Discharge Data; HAE is Hospital Adverse Events 13 X
Analysis of Nurse Staffing Options Option 1: Nurse staffing for entire hospital (both inpatient and outpatient) using HAR data Option 2: First Quarter 2014 MHA staffing plan data (no skill mix), unit type (level of granularity at this point is unclear) Option 3: Retrospective years using MHA staffing plan data as a benchmark adjusting by HAR skill mix data 14
MEDIUM TERM APPROACH July October 2014 15
Part 1: Description of Nurse Staffing Variation in Minnesota MHA Nurse Staffing Report (July 2014) Inpatient nurse staff only Non-managerial staff direct patient care Staffing by unit type: ICU, non-icu types 16
Part 2: Nurse Staffing & Patient Outcomes: Important Related Factors Patient factors: Hospital Discharge Data (CY 2013 and prior) Age, sex, patient acuity (APR-DRG severity score, comorbidities) primary payer, ZIP Hospital factors: Hospital Discharge Data (CY 2013 and prior) Case mix index for all hospitals HAR (FY 2012) and Other Data Sources: Teaching status, urban-rural location, critical access, bed size, ownership, total admissions, and payer mix Medicare case mix index for PPS hospitals (FY 2013) 17
Part 2: Nurse Staffing & Patient Outcomes Analytical Options Option 1: Analyze only concurrent data Nurse Staffing MHA Nurse Staffing Reports 1 st Qrt. 2014 Patient Outcomes Collect HDD data from hospitals 1 st Qrt. 2014 Other Factors HDD data from hospitals 1 st Qrt. 2014 Option 2: Analyze historical data Nurse Staffing MHA Nurse Staffing Reports 1 st Qrt. 2014 Patient Outcomes Historical HAE, HDD (FY 2010-2013) Other Factors Historical HDD (CY 2010-2013), other sources Option 3: Analyze historical data using additional data Nurse Staffing Collect historical data w/ skill mix (FY 2010-2013) Patient Outcomes Historical HAE, HDD (FY 2010-2013) Other Factors HDD, other sources (FY 2010-2013) 18
POTENTIAL LONG-TERM APPROACH Post January 2015 19
More Robust Studies Would Require Currently Unavailable Data Nurse experience Organizational culture Physical space and layout Available technologies (both patient care and administrative) Organizational structure and hierarchy Administrative practices, including use of float nurses Available supports and specialties (such as wound care consults, therapy teams, transport teams) Use of rapid response teams Work environment Separation of staffing in non-icu hospital unit types 20
DISCUSSION 21
Questions for Discussion Where to focus on outcome measures? How to arrive at alignment of data on outcomes, staffing and claims over time? Where should MDH focus efforts in further data collection, if at all? What preliminary and final analyses should MDH consider? Other comments? 22
Contact Information Nate Hierlmaier (651)201-3541 nathan.hierlmaier@state.mn.us Stefan Gildemeister (651) 201-3550 stefan.gildemeister@state.mn.us www.health.state.mn.us/healtheconomics 23