Introduction

This concept discusses the analysis of waiting times to surgery in studies completed at the Manitoba Centre for Health Policy (MCHP). It identifies which surgical procedures have been evaluated, what registries are used to gather information to determine wait times, how wait times are defined, statistical analysis that has been used to evaluate waiting times, and diagnostic, procedure and tariff codes used to identify surgeries and other events.

Background

Waiting times were first defined as the time between the date of a patient's last pre-operative visit to the surgeon and the date of the surgery (De Coster et al., 1998). This measure of surgical waiting time was developed because, except for cataract, cardiac, and joint replacement surgery, there was no centralized surgery registry in Manitoba. And, in using administrative data (routinely collected computerized hospital discharge abstracts data and medical services/physician claims data) to estimate the wait for surgery, we had no field indicating when the patient and the physician made a decision to proceed with surgery. This method is still used for surgeries where these is not a centralized wait time registry.

This method assumes that the patient is referred to a surgeon for a particular problem, that the surgeon and patient decide to proceed with surgery (beginning of wait), after which any problems that arise are handled by the patient's family physician, and the patient doesn't see the surgeon again until the date of the surgery (end of wait): in short, that there is usually only one pre-operative visit to the surgeon. This method was similar to one used by the Nova Scotia Department of Health (1996).


• This method isn't suitable for procedures for chronic conditions (e.g. total joint replacement, hysterectomy for benign disease) that might involve a number of pre-surgical visits to the surgeon.
  
• This method estimates waiting times only for patients who had the procedure; persons who were going to have the surgery but did not - because they left the province, died, changed their minds or were still waiting - were not in the data.

Registries

Surgical waitlist registries provide complementary and additional information on surgical cases beyond what is available in the administrative data. Cataract, cardiac and joint replacement (hip and knee) surgeries in Manitoba are captured in the three following registries that are available at MCHP:


1. Manitoba Cataract Surgery Waiting List Program (MCSWLP) Registry
   
   • Includes patients who are approved for and have agreed to undergo cataract surgery in Winnipeg (90% of all cases in Manitoba).
   • Patient information is archived when they receive surgery, or the surgery is cancelled.
   • De Coster et al. (2007) used archived data to estimate wait times for cataract surgery.
   • Please see the Manitoba Cataract Surgery Waiting List Program Database for more information.
   
   
2. Cardiaccess Registry
   
   • Includes patients who are approved for and have agreed to undergo cardiac surgery in Winnipeg (100% of all cardiac surgeries)
   • Please see the Cardiaccess Registry Database for more information
   
   
3. Joint Replacement Waiting List Registry
   
   • Includes patients who have been approved and have agreed to undergo hip/knee replacements (HKR) surgery in Winnipeg (90-94% of all HKR in Manitoba)
   • Voluntary registration of patients until 2005
   • Please see the Joint Replacement Waiting List Registry Database for more information

Patients are entered into these registries when they decide to undergo surgery. Registries that have voluntary registration (i.e. HKR Registry until 2005), will not capture all surgical cases.

De Coster et al. (2007) linked the registries with administrative data through encrypted PHINs, date of surgery and surgical procedure codes to:
• "compare the two data sources to assess validity of wait time estimates from administrative data" and
• "to incorporate clinical and functional data in the multivariate models to determine which variables are related to variation in wait times" ( p.35).

Records without an encrypted PHIN were excluded by De Coster et al. (2007).

Methods to Estimate Waiting Times

These methods have been used by De Coster et al. (1998) and De Coster et al. (2007) to estimate waiting times to surgery for the surgeries listed above.

A) Procedures Investigated

1) De Coster et al. (1998) examined eleven procedures :

• elective surgical procedures
  
  • cholecystectomy
  • hernia repair
  • excision of breast lesions
  • varicose vein repair
  • carpal tunnel release
  • trans-urethral prostatectomy
  • tonsillectomy
  • carotid endarterectomy
• cataract surgery
  
• coronary procedures:
  
  • coronary artery bypass surgery (CABS)
  • coronary angioplasty (PTCA)

2) De Coster et al. (2007) analyzed:

• elective surgeries (listed above), with adenoidectomy in combination with tonsillectomy
  
• cataract surgery
  
• cardiac surgery
  
  • coronary artery bypass graft (CABG)
    
  • heart value replacement
• joint replacement surgery
  
  • primary hip replacement
  • primary knee replacement

B) Inclusion/Exclusion Criteria

• Only procedures that were the primary reason for hospitalization (indicated by the procedure code being in the first position) were included (De Coster et al., 1998; De Coster et al. 2007).
  
• Patients whose visit to the surgeon was less than three days pre-op and less than four days pre-op were excluded by De Coster et al. (1998) and De Coster et al. (2007), respectively, (except for coronary patients).
  
• Elective/day procedures were included in the analysis of all surgeries. Only coronary surgery cases included urgent/emergent cases as well. The urgent coronary surgeries were analyzed separately from elective surgeries (De Coster et al.,1998; De Coster et al., 2007).
  
• If a patient had more than one surgical procedure within the years of analysis, only the first procedure was included (second, third, etc. surgical procedures of that patient were excluded) (De Coster et al., 1998; De Coster et al., 2007).
  
• Patients without pre-op visits were excluded (De Coster et al., 1998).
  
• Out-of-province patients were excluded (De Coster et al., 1998, De Coster et al., 2007).

Figure 1: Records used to estimate elective surgery waiting times in De Coster et al. (1998) gives a pictographic description of inclusion and exclusion criteria for elective surgery cases.

C) Algorithms

The following are various definitions of waiting times for elective, cataract, cardiac and joint replacement used in De Coster et al. (1998) and De Coster et al. (2007) .

1) De Coster et al. (1998)

• i) Elective Surgeries
  
  • Waiting time = time between the pre-operative visit to the operating surgeon and the date of surgery.
  
  
• ii) Cataract Surgeries
  
  • Waiting time = time between the pre-operative visit to the operating surgeon and the surgical date.
    • If the pro-operative visit was for an axial measurement of the eye (ultrasound), the next closest date was used as the start of the wait time.
  • No "three day" rule was used; cases in which the patient had their pre-operative visit within three days of the surgery were included.
  
  
  During the period of this study, cataract surgery in Manitoba was available in only the cities of Brandon and Winnipeg.
  
  • Both cities had private clinics as well as public hospitals performing the procedure.
    
  • In both cases, the surgeon's fee was paid by Manitoba Health; however private clinics charged a "facility" fee of about 1,000 dollars which was not covered by Manitoba Health.
    
  • Waiting times were shorter for cataract surgery done in a private clinic compared to in the public hospital.
    
  • When cataract patients were categorized according to where their surgeon practiced - those who operated publicly only vs. those who operated both publicly and privately - the waiting time picture was different.

    Note : no surgeons only operated privately

  • Patients who had cataract surgery in a public hospital by a surgeon who operated publicly had to wait from 7 to 10 weeks.
    
  • However, if the surgeon operated both privately and publicly their public hospital patients waited from 14 to 23 weeks. Meanwhile their private clinic patients had to wait, on average, about 4 weeks.
  
  Note : the surgeons who operated both publicly and privately performed over 50% of the cataract surgery provided in the public sector; it is possible that the longer waits for public sector surgery reflected the limits of available resources.
  
  
• iii) Coronary Surgeries
  
  • Coronary Artery Bypass Surgery (CABS)
    
    • CABS requires an angiogram and a surgical consult prior to surgery
    • Waiting time = time between the angiogram or surgical consult (using the one closest to the surgery) and the surgery date.
      • If there is more than 1 pair of angiograms and surgical consults, the pair closest to the surgery is used.
    
    
  • Percutaneous Transluminal Coronary Angioplasty (PTCA)
    
    • Waiting time = time between the last angiogram and surgical date (angioplasty)
    • PTCA is not a surgical procedure so there is no pre-op visit to a surgeon; therefore angiography was used as the marker for waiting time for PTCA.

2) De Coster et al. (2007)

• i) Elective Surgeries
  
  • Waiting times were measured with the same method as in De Coster et al. (1998).
  
  
• ii) Cataract Surgeries
  Used algorithm developed by De Coster et al. (2002):
  
  • Waiting time = time between pre-op visit (see below) and the surgical date.
    • If there was only 1 pre-op visit, then this visit was used as the marker for the start of the wait time.
    • If the pre-op visit was an axial measurement of the eye (tariff code 9890, 9891), then the second closest visit was used as the marker for the start of the wait time.
    • If the pre-op visit was within 70 days of surgery, then the second closest visit was used as the marker for the start of the wait time.
    • If the second closest visit was for radiology, the third closest visit was used as the marker for the start of the wait time.
  
  
• iii) Coronary Surgeries
  
  • Coronary Artery Bypass Surgery (CABS) - Elective
    
    • Waiting time = time between closest pre-operative visit (assuming it was more than 3 days before the surgery) and surgery date
      • If there was no pre-op visit, the date of an angiogram was used as the start of the wait time.
    
    
  • Coronary Artery Bypass Surgery (CABS) - Urgent
    
    • Waiting time = time between the closest angiogram and the surgery date
    
    
  • Valve Replacement
    
    • Waiting time = time between closest pre-operative visit and surgery date
      • If the closest visit was within 28 days of the surgery, the second closest visit was used as the marker for the start of the wait time.
    • Only used elective surgeries
  
  
• iv) Joint Replacement Surgeries
  
  • Unable to develop an algorithm to estimate waiting times for HKR
  • De Coster et al. (2007) reported that there was no relationship between the start date in the registry and the date of a pre-surgical visit.

D) Analysis of Waiting Times

The following is a description of how wait times were analyzed by De Coster et al. (1998) and De Coster et al. (2007) by calculating medians and means, and constructing box-plots and multivariate models.

1) De Coster et al. (1998)

Waiting times were estimated for each year for 5 year periods (from 92/93 to 96/97, except for the coronary procedures in which 1990/1991 to 1996/1997 were used.

i) Calculating Average Waiting Times: Median vs. Mean


Both median waiting times (the mid-point, the length of time half the people overa given time period had to wait for surgery) and mean waiting times (see Confidence Interval of Median and Statistical Analysis of Mean Waiting Times concepts) were used.


• The median, unlike the mean, is not influenced by outliers (i.e. unusually short or long waiting times).
  
• 95% confidence intervals were calculated for all comparisons, with adjustment for multiple comparisons.
  
• Programming code for calculating confidence intervals for the median is available.
  
• One limitation of median analyses was that it doesn't allow adjustments for other factors in making comparisons.
  
• Mean analysis resolves this problem and is statistically more powerful and sensitive.


In order to deal with the problem of outliers, Tukey's robust outlier detection method was used to separate a few extremely unusual waiting times.


• The interquartile range (IQR) was calculated first.
  
• Outliers were identified as waiting times which were longer than 3 times the IQR + the 75th percentile or shorter than the 25th percentile - [3*IQR]. Outliers were excluded from all mean analyses thereby assuring a robust comparison of the mean waiting times, uninfluenced by a few extreme values.

Based on the analysis of medians and substantive considerations, a decision was made to take mean waiting times of longer than three days as a clinically significant characteristic; no average waiting times less than this were tested for statistical significance. (see DeCoster et al (1998), page 71).

2) De Coster et al. (2007)

Waiting times were estimated for the years 1999/2000 to 2003/2004.

i) Calculating Unadjusted Wait Times: Box-Plots


Box-plots were used by De Coster et al. (2007) to analyze the variation in the distribution of waiting times by year and by Regional Health Authority (RHA) of residence of patient. The box-plots display the waiting time to surgery by the 10th, 25th, 50th, 75th and 90th percentiles. The percentiles are explained as the length of time it took for X% of patients to receive their surgery (e.g., the wait time at the 10th percentile indicates how long it look for 10% of the patients to receive the surgery).


• The median wait time from the box-plot analyses (50th percentile) should not be compared across years or RHAs because "the patients being treated within an RHA or within a year may be fundamentally different from the patients being treated within another RHA or year" (De Coster et al., 2007, p.6).
• This issue is solved by adjusting median wait times "so that each RHA or year is 'equal' on all other important factors that may influence wait times" (p.6)

ii) Calculating Adjusted Median Wait Times: Multivariate Models


The median wait times were adjusted by De Coster et al. (2007) using multivariate models to allow comparison across years or RHAs and to estimate the factors associated with wait time variation. Data from 1997/1998 and 1998/1999 were added to the survival analysis for baseline comparisons. Variables from administrative data and registries were included in the multivariate models through a two step process:


1. All administrative variables were included
2. Variables of interest from the registry were added to the models (pattern mixture model)

Unfortunately, a large number of variables were missing from records in the registry. Logistic regressions were completed to determine whether the variables were missing systematically or randomly. Missing variables must be included in the multivariate models to eliminate bias, but they cannot be interpreted. The purpose of step 2 - pattern mixture model (see above) is to allow unbiased estimates to be completed of how variables affect outcomes of interest (wait time in this case). Step 2 was not required in the cataract multivariate model because the MCSWLP Registry is 100% complete for the years studied (1999/2000-2003/2004).

Table 5: Variables used in models for cataract, cardiac, and HKR surgery identifies variables from administrative data and the registries included in the multivariate models.

In general, waiting time distributions "rise quickly and stretch out to the right" (De Coster et al., 2007, p.6). De Coster et al. (2007) cautions that "methods that assume a normal distribution, such as ordinary least squares regression, or other methods based on the general linear model, should not be employed" (p.6). Parametric survival analysis is appropriate to use for this type of data and is intended to evaluate the time to the occurrence of an event, in this case, surgery.

To directly compare the wait times across year or RHA, an adjusted wait time curve can be created. Please see p.8 for an example figure of an Adjusted Wait Time Curve . This graph will display the adjusted median wait times. De Coster et al. (2007) presented the adjusted median wait times in bar graphs by year and by RHA, with 2003/2004 and Winnipeg as its referent, respectively.

E) Diagnostic, Procedure and Physician Tariff Codes

Diagnostic codes, procedure codes, and physician tariff codes are available from MCHP. Appendix 1 in De Coster et al. (2007) provides tables identifying the tariff codes and ICD-9-CM procedure and diagnostic codes for the eight elective procedures and cataract, cardiac and joint replacement surgeries that were investigated in this project.