The authors describe the develop- ment of a model which assists with answering the question, “for a facil- ity with certain characteristics, what would be the industry average num- ber of security FTE’s?” Through the use of multiple regression analysis using actual field data, an objective means of making this determination is possible, they report.

(Karim H. Vellani, CPP, CSC, is the President of Threat Analysis Group, LLC, Sugar Land, TX, an independent security consulting firm. He is the author of two books, Applied Crime Analysis and Strategic Security Management, and has contributed to a number of other secu- rity related books and journals. He is a mem- ber of IAHSS.

Dr. Robert Emery, PhD, CHP, CIH, CBSP, CSP, CHMM, CPP, ARM is Vice President for Safety, Health, Environment & Risk Manage- ment for The University of Texas Health Sci- ence Center at Houston and Professor of Occupational Health at the University of Texas School of Public Health. He possesses national board certification in all of the main areas of health & safety.

Nathan Parker is a researcher at the University of Texas School of Public Health. He holds a Bachelor of Science degree in biology from Lafayette College and a Master of Public Health degree in epidemiology from the Uni- versity of Texas School of Public Health.)

Because of the unique rolethey fulfill within their com- munities, hospitals often serve as a hub of human activity. In the U.S., hospitals employ over five million individuals, making it the second largest source of private sector jobs in the country (AHA 2012). In 2008, hospitals in the U.S. treated 123 million people in their emergency departments, provided care for 624 million out- patients, performed 27 million surgeries, and delivered four mil- lion babies (AHA 2012). Unfor- tunately, accompanying this massive amount of human activ- ity is a variety of security risks. Recent government reports also indicate that both the frequency and severity of criminal activities occurring within hospitals is on the rise or as the Joint Commis- sion stated, “Once considered safe havens, health care institu- tions today are confronting


Staffing benchmarks: a model for determining how many security officers are enough Karim H. Vellani, CPP, CSC, Robert J. Emery PhD, CHP, CIH, CBSP, CSP, CHMM, CPP, ARM and Nathan Parker

steadily increasing rates of crime, including violent crimes such as assault, rape and homicide” (Joint Commission 2010). Although se- curity risks can vary from hospital to hospital and from region to re- gion, it becomes the job of the hospital executives to determine the amount of resources that must to be diverted from the primary mission of patient care to support the security function.

The challenge of determining the amount of security resources that are appropriate for any par- ticular organization is made par- ticularly difficult because the security profession functions largely in the realm of prevention --put simply: “on a good day, nothing bad happens.” So the question becomes “what amount of resources is needed to make nothing happen?”


Hospitals have individual secu- rity needs and face unique secu- rity risks. Comparing one hospital to another, even to those of similar size and in the same ge- ographic area, illustrates the dis-

parity in security measures. Se- curity staffing levels, in particular, can vary dramatically from hos- pital to hospital. The need for boots on the ground is easy to jus- tify. The number of boots needed, on the other hand, is the real chal- lenge and one that becomes in- creasingly more difficult each budget season. Like all security measures and prevention efforts, the value of and return on invest- ment from security officers is dif- ficult if not impossible to quantify on a risk basis. And in fact, sur- rogate factors are often cited as performance metrics. For exam- ple, customer service levels or pa- trol frequency are surrogate factors that may be used as met- rics to judge performance. Hos- pitals also benchmark their security programs against other hospitals. How often do we draw comparisons such as “Hospital X has metal detectors and so should we.”

What works at one hospital may not work at another. Cultural fit, cost, and convenience must be taken into account when deter- mining what security measures are appropriate for a given hospi- tal. However, there are some se- curity measures that are worth


and benchmarking. Security staffing levels is one such meas- ure and is addressed in this article. Because hospitals have unique se- curity risks, this study does not at- tempt to identify the appropriate security staffing level for a partic- ular hospital. Alternatively, our goal is to identify general industry benchmarks for hospital security staffing using a data driven ap- proach.


In the hospital environment, se- curity officers can serve diverse functions depending on the hos- pital’s needs and the Security Di- rector’s willingness to take on responsibilities that may not be within the traditional scope of work for a security department. In a non-traditional security envi- ronment, security officer duties can range from patrol and re- sponse, visitor management, pa- tient observations, bank deposit deliveries, visitor and patient es- corts, and parking assistance and enforcement. With few excep- tions, the one consistent duty for Security Officers is patrol and re-

sponse. Despite the fact that the patrol

and response duty is the one con- sistent function of a security offi- cer, it is also the one duty in which staffing is often based on an inad- equate number of drivers. Many healthcare consultants and insid- ers use square footage as the driv- ing force for determining the number of security officers needed. Others use security call volume. And yet others use some other factor. Our research, and common sense, indicates that the use of a singular factor is insuffi- cient.

The notion of quantifying the number of security officer needed is somewhat akin to the way the price is set for a used car. If asked what a particular used car is worth, most people would re- spond that the amount they would pay depends on aspects such as the year, make, model, mileage, condition of the vehicle, color, ad- ditional features, and so on. But if one actually collects a suitable amount of market data, one can determine that there exists a finite group of key predictors that can be used to reliably predict the value of a used car. The key term here is “group” as it is usually not


a single predictor, but a combina- tion of several predictors that al- lows this projection. In the case of a used car, the likely group of key predictors includes make, model, mileage, general condi- tion. Using this same approach, one can begin to estimate or pre- dict the industry average number of security staff for a particular hospital. The challenge is first identifying the key predictors, and then assembling the market data. Addressing these two challenges is part science and part art, but the information that can be produced from this exercise can be very useful.

Identifying The Key Predictors The first step in the develop-

ment of the envisioned security staffing model was to bring to- gether a group of hospital security professionals to gather their in- sights on what characteristics and parameters drive the need for se- curity staffing levels in the hospi- tal environment. Given that the Texas Medical Center (in Hous- ton, Texas) is the world’s largest medical center, the assembly of such a group was not terribly dif- ficult. The group members in- cluded: Geoff Povinelli, Director, Security Services, The Methodist

Hospital; Bert Gumeringer, Director, Facilities Operations & Security Services, Texas Chil- dren’s Hospital; Joe Bellino, System Executive, Integrated Protective Services, Memorial Hermann Healthcare System; Meco Choates, (former) Director, Public Safety, The Methodist Hospital.

During the initial meeting, an open discussion was held in an at- tempt to list the universe of as- pects and characteristics that could drive the need for security in a hospital. This list is shown in Figure 1.

Once created, the next step was to determine which of these char- acteristics or aspects were both quantifiable and readily obtain- able. The notion of possible “in- teraction” was also considered. For example, the generally ac- cepted way to convey the size of a hospital is through “bed counts.” But size can also be measured by parameters such as “total inpatient clinical square footage”, so an examination is needed to determine which pa- rameters might serve as surro- gates for the others and where statistical interactions might occur.


FIGURE 1 Aspects and Characteristics of Hospitals That May Drive the Need for

Security 1. Size indicators

a. Number of Licensed Beds b. Total square footage that is covered by security program

(officers or systems) c. Inpatient Clinical square footage d. Outpatient Clinical square footage e. Research square footage f. Administrative t square footage g. Exterior square footage (grounds, common areas, and parking) h. Construction square footage

2. Risk indicators a. Security Call Volume b. Area Crime Statistics c. Campus Security Incidents d. High Risk Areas (ED, Trauma Center, Psych Unit, L&D) e. Area Demographics and Cost of Living f. Patient Demographics g. Payer Mix (Ratio of Public Assistance to Private Insurance) h. Unique Departments (Psychiatric/Behavioral Health,

Pediatrics, Geriatric, Drug/Alcohol, Trauma Center, Nuclear Medicine, Medical Research)

i. Campus Complexity (e.g. number of entrances open day and night)

j. Geographic Area k. Urban v. Suburban v. Rural

3. Population/Traffic indicators a. Clinical Staffing Level b. Admin and Support Staffing Level c. Patient Census d. Visitor Traffic Level e. Emergency Department Visits f. Adjusted Patient Discharge/Days


Determining The Top Five Predictor Variables of FTE's

It was through subsequent dis- cussions where the subset list was derived in Figure 1. Based on this consensus list, the pilot group par- ticipants agreed to return to their respective institutions and gather this information for compilation and analysis. Armed with the data from 15 participating hospi- tals during the first pilot effort, statistical assessments were made based on reported full time equiv- alents1 (FTE). The assessment indicated that the top predictor variables were:

1.Total inpatient clinical square footage

2.Total annual number of security calls

3.Total research square footage 2

4.Total number of hospital beds 5.Presence of trauma center The predictors seemed intuitive

for predicting staffing levels. The survey results and analysis was then re-examined by the focus group. Given the small sample size, the potential for misleading statistical inferences existed, so the consensus opinion was to at- tempt to collect are larger set of data, based on the previously

identified likely key predictors along with several others that might be useful. Armed with a working model, we were able to fairly accurately predict how many security officers were in place at a given hospital by iden- tifying answers to the predictors above. More importantly, we were able to narrow the list of sur- vey questions for a second, larger pilot. In November 2011, a re- vised survey was distributed elec- tronically using Survey Monkey to the membership of the Interna- tional Association of Healthcare Safety and Security (IAHSS), asking for eight items:

1.Number of FTE’s 3 assigned to: a.fixed posts b.patrol/response functions

2.Total interior square footage 3.Number of licensed hospital

beds 1 For the first pilot, we defined FTE as the number of full-time equivalents authorized by hospital security department to work 40 hours per week, including contract and proprietary off-duty police, unarmed security officers, armed security officers, dispatchers, and operators. 2 Note that medical research is a common activity in the Texas Medical Center and as hypothesized, research space was a top predictor. 3 For the second pilot, we defined FTE as the num- ber of full-time equivalents authorized by hospital security department to work 40 hours per week, in- cluding contract and proprietary off-duty police, un- armed security officers, and armed security officers. We excluded dispatchers and operators this time.



R egression A

nalysis of R eported H

ospital Security F


’s and V arious P


4.Total inpatient clinical square footage

5.Total research square footage 6.Total number of security calls

in 2011 7.Presence of a trauma center

(yes or no, and what level) 8.Presence of a specialist unit

such as: a.Psychiatric/behavioral health b.Pharmacy c.Retail pharmacy d.Labor and delivery e.Emergency department f.Nuclear medicine g.Medical records department

The Revised Set of Top Predictors Of Total Security FTE's

The second pilot yielded 94 re- sponses to the revised survey by January 2012. This pilot also un- covered an issue in predicting the number of FTE’s dedicated to fixed posts for two reasons. First, many hospitals do not have fixed posts as was the case with our sur- vey respondents. Second, fixed posts vary by hospital depending on several factors including spe- cialized posts, visitor manage- ment, reception functions, etc. So our focus shifted from total FTE’s in the first pilot to total FTE’s and

total FTE’s dedicated to patrol and response in the second pilot.

The regression analysis 4 of the 94 responses indicated a somewhat revised set of top five predictors of total security FTE’s:

1.Presence of psychiatric or behavioral health unit

2.Total annual number of security calls

3.Presence of level I trauma center

4.Number of hospital beds 5.Presence of level IV

trauma center We are now able to predict the

total FTE’s at a hospital with these predictors using the follow- ing formula:

Total FTEs = e^[0.388704 (Psych/behavioral health unit) + 2.32x10 (Security call volume) + 0.6285178 (Level one trauma center)+ 0.0002007 (Hospital beds) – 1.207758(Level four trauma center) + 2.602291]

4 “In statistics, regression analysis includes many techniques for modeling and analyzing several variables, when the focus is on the relationship be- tween a dependent variable and one or more inde- pendent variables. More specifically, regression analysis helps one understand how the typical value of the dependent variable changes when any one of the independent variables is varied, while the other independent variables are held fixed” (



R egression A

nalysis of R eported H

ospital Security F T

E ’s

A ssigned to P

atrol and R esponse and V

arious P redictors

As expected, the top predictor variables for identifying the FTE’s dedicated to patrol and re- sponse is similar to the variables for predicting total FTE’s:

1.Total security calls 2.Total hospital beds 3.Presence of level I trauma

center 4.Presence of a psychiatric/

behavioral unit Using a slightly different for-

mula, we can also predict the FTE’s dedicated to patrol and re- sponse:

Total FTEs = [7.65x10-6 (Secu- rity call volume) + 0.0006511 (Hospital beds) + 1.5773 (Level one trauma center) + 1.0259 (Psych/behavioral health unit) + 2.69056]2

Square Footage No Longer A Significant Predictor

As seen in the list of predictors, square footage was no longer a significant predictor as com- monly believed. Using the up- dated model, our ad hoc tests have proven successful when pre- dicting staffing levels at hospitals including those that that did not respond to the first or second sur- vey. If a hospital Security Direc- tor provides answers to the four

predictor questions above, we can now provide them with a pre- dicted staffing level for FTE’s dedicated to the patrol and re- sponse function. For example, in a hospital with 100,000 security calls, 600 beds, a Level 1 trauma center, and a psychiatric unit, the model tells us that the benchmark is 41.6 FTE’s dedicated to the pa- trol and response function.


However, it is important to un- derstand the limitations of any method used for predicting “in- dustry averages.” The model de- veloped thus far from this exercise merely assists with an- swering the question, “for a facil- ity with certain characteristics, what would be the industry aver- age number of security FTE’s?” There are many seasoned security professionals who could rely on their knowledge and experience to come up with a number, but through the use of multiple re- gression analysis using actual field data, a more objective means of making this determination is possible. The information repre- sented by the model could be of tremendous use to hospital exec-


utives who are considering ad- justing the security staffing levels at their organizations. It is also important to underscore that the model in no way speaks either the performance of the security force or to security outcomes.

While the model has developed significantly over the past two years, this is not the end of the analysis. The model is a guide and nothing more. Unique factors at each hospital may drive staffing levels up or down. Moreover, the model will likely change as we collect more data.

The next step for this project is to develop a third pilot by revising the survey based on feedback re-

ceived at the 2012 IAHSS AGM and gathering additional survey responses. The more responses received, the more accurate the model becomes. So when you see the email requesting your help with the survey, please respond. In the meantime, if you’d like to know what the model says for your hospital, feel free to email Karim Vellani at [email protected] References

American Hospital Association initiatives/econcontributors.shtml

Joint Commission h t t p : / / w w w. j o i n t c o m m i s s i o n . o r g / assets/1/18/SEA_45.PDF