The Computer as an Unbiased Medical Investigator Experience on an Active Surgical Service Richard M. Stillman, MD, Brooklyn, New York Winston G. Mitchell, MD, Brooklyn, New York Gerald W. Shaftan, MD, FACS, Brooklyn, New York Philip N. Sawyer, MD, FACS, Brooklyn, New York Samuel L. Kountz, MD, FACS, Brooklyn, New York
With the expanding complexity of diagnostic and therapeutic modalities, the difficulty of performing an accurate retrospective study is increasing and hence vital decisions are often based on studies of relatively few patients while data on thousands of similar cases go unrecorded. Anyone who has ever attempted to complete a sizable chart review is altogether too familiar with the inevitable problems of handwritten, disorganized medical records. It is similarly frustrating to the physician to go over the chart of readmitted patients for it is often impossible to identify and interpret significant information. Incomplete, illegible or erroneous data, the unavoidable bias of the reviewer, and the prohibitive time required for an accurate study are responsible for distortion or loss of valuable clinical data. Each physician will have knowledge of his personal cases but will not have an overall appreciation of the results obtained by others. Ideally, the raw data for the retrospective study should be compiled prospectively by the physician responsible for the care of each patient. These problems in medical data processing now demand investigation and solution in order to allow implementation of the Professional Standard Review Organization (PSRO), which became a fact by Public Law 92-603. Assurance of high quality health care requires facilities for screening numerous parameters in each of many hospital admissions extending from concurrent review (while the patient is still hospitalized) through individual case review to retrospective pattern-of-care review [I].
From the Department of Surgery, State Univw~lly ofNew York, Downstate Medical Cater, and Kings County HostSSalCenter, Brocklyn, New York. This work was supported in part by the SNYMERF Fud. Reprint requests should be addressed to Richard M. Stillman. MD, Department of Surgery, Downstate Medical Center, Brooklyn, New York 11203. Presented at the Forty-Seventh Annual Meeting of the Pacific Coast Surgical Association, Monterey, California. February 15-18, 1976.
Volume 132, August 1976
For these reasons computerized medical charting has received increasing attention, and numerous systems have been designed to handle this formidable problem [Z-4]. Although used satisfactorily by their originators, most of the previous computer systems find little application outside their place of origin due to the complexity of programming for other specialities or difficulty of application in the hands of inexperienced users. In addition, the magnitude of the problem has required acceptance of often too strict limitations in the scope of information desired. We believe that the key to the solution of these problems lies in the discharge summary. While medical records as a whole document the detailed course of a patient’s hospitalization, discharge summaries should attempt to condense the material present on the chart without omitting major facts. Presented here is our experience withan adaptable computerized medical data system (MEDS) which was designed to assure not only simplicity of entry of the discharge summary and accuracy and speed in performing statistical analyses from the information stored, but also ease of actually programming the computer for eliciting the information required on any particular service [5J. Methods In a room adjacent to the Trauma Surgery Ward at Kings County Hospital there is a Hazeltine 2000 video screen terminal and printer (similar to that used by airlines at their ticket counters). This is connected by ordinary telephone lines to a large PDP-10 digital computer 40 miles away. This machine is available 24 hours a day for entry, retrieval, and analysis of discharge summaries. On the wall above the terminal is posted a short instruction sheet showing the user how to turn on the terminal, dial the computer, and initiate communication. On admission, the biographical data for each patient are entered into the computer by the secretary. On discharge,
209
Stillman et al CIHCUMSlANCE5 OF INJUicY I ASSAULT 2 MGTOh VEHICLE 3 ANIMAL ATTACK 4 SUICIDAL 5 FALL 0 MACHINEHY I SPOHTS 8 F.B.INGEST. 9 hOME ACCIOENT 10 OTHtH CH(X)SE ALL AI’PLICABLE 7 _I_
. .._ _. ..-....__... . .-
AGEhT I SHAHP INSTWUMENT 2 BULLET 3 SHUT GUN 4 BLUNT TIlAUKA 5 HUllAN BITE 6 ANIMAL BITE I MOTOR VEHICLE B F.B. 9 OTHER CHOOSE ALL APPLICABLE ? 2 BULLET (CALIBRE) = 7 & THAT SEEMS HATHEH HIGH. ARE YOU CEKTAIN BULLET IS 83 CALIBWE (YES OH NO) ? w BULLET (CALIBRE) = ? a
AREA(S) INVOLVED I HEAU 2 NECK 3 VERTtWHAE 4 CHEST 5 ABGOYEN 6 EXTREMITY I GYk U YELVIS I(T. 9 PELVIS LT. CH(X)SE ALL APPLICABLE
1HAT
_
/MISPITAL
CWWSE/
CONCOYlTrANT UIItASEI *ONE ANT‘B,OT,C “SAGE, PWWHYLACIIC, ,YH,GAr,ON OUTCOHEI COWLICATED CWPL.8 PULWNAHY (UN M)SP,TAL DAY I21 P,,LYONA,,Y U,YPLICATIO~S~ ATkLtCTh 15 “,SCkA”OE COND,T,ONI UISCHAWEU IHPRWE”
72
/ABDOMEN/
SIGNS
4,
Figure 1. Portion of a sampfe Interactfon during lfw 5 fo 10 minute Process of compllhtg a discharge summery for a surgical patlent. Users responses are under&ml for clarrtv.
entry by the physician
SLOW
lOStal *“LLEr 138 CALIRML) ENTnY WOUNDI “W EXlT 1)0UN08 SACK SYWTMS8 PAIN - LOCAL1 IZED LOCAL,,?E” O”A”“,NO. AHNOPMAL S”1EL SWWS SIGNS, LOCALILE” lENbE cNtSS. STUD,651 *SDw.Y)r*“. “I 1INALYbIS URINALYSISI NORMAL *BDOI,NAL WAY kES”L,l NORMAL AHEAS OF ASU.,NVOLVEDI ARD.NALL. DEF,NIrC ‘“TDA-aL)” ‘,A’ T,,EYACYl ,MMUIE”,ATE OYtWATION TlYt SElllEEY ADY,SS,Ok AhD SUltGE)tY* I-2 MS. TYPE ““E-W HEPLACEMENT: ll,!OLE RLWO (2 “NITS), CHYSTALLOI us oom cc, TYPE 1NT”A-W WEPLACtYENTl *HOLE BLOW L3 “NITS,. PLASMA , 250 cc,. CHfSTALLOIUS ti !ocm cc) .“* “F”IWNEAL P1LNET”ATIW. RLW” IN PEYITOhhAL CAVITY, FIkUlkGS AT L... ._ . . SO, I,, WOAN IN.,“PV OWCiirNm LIVEN LlVEk LO&Em ulGH1 LIVE” ,N.,U”Yl PUNCTUW’ ( SU LLET 1 A, TIME OF EYPLOWATTON LlVE” ,NJUHYI SLEEDlhG ~~X.L,YEHI smwt ~16.. “YA,N(S,USY”* SUMP AhD %%k’ NOM “EGHEt OF ARUOYINAL CM “TTAL,NAT,,,N*
_..___ ._.. -...
B ABlrOhMAL BONEL SOUNDS 9 OMENTAL HELENI AT ION IO BOkEL HEkNIATION II MASS 12 NOM OF TkSE CH(X)SE ALL APPLICABLE 7 2.
I
/AB”ONEN,
SYMPTOMS I PAIN - LOCAL I ZED 2 PAIN - GENEdALIZED 3 I’AIN - RiiFt~MEll TO SHOULDER 4 PAIN - RLFEtMED TO FACK :, HEMATEMES IS b VOMI’IING 7 NAUSEA 8 CONS1 IPATIOti 9 HtLENA IO NOnb OF TkSE CH(X)SE ALL APPLICARLE 7L
of the patient’s name or chart number provides the key to identification of the discharge summary and initiates a series of questions that appear on the screen. (Figure 1.) As the answer to each of these is typed, the machine checks it for validity and saves the response in a coded form. When the interaction is complete, a copy of the discharge summary is printed (Figure 2) for signature and insertion into the patient’s hospital chart. Confidentiality is guaranteed because a secret password is required to gain access to the stored information.
210
..~
TRAUMATIC@ EYEWENCY TlYE FROM INJUWY TO APWIVAL AT HOSPITAL (HOUWS,I UP TO CONDITION ON ADY,SS,OH* SHKX VITAL S,ONS YEWYUE” ON ADY,SS,ONr PULSE ,120 SEATS/YIN), PRESSUWE SYSTOLIC (00 MY/%). S.P.U,ASTOL,C (3” MWHG) RESUSCI’TATION‘ RLW3. CLEAV FLUID “ESYWSS TO WESUSCITATION~ GCOD CIUCUllSTANCES W lNJLl”Yr ASSAULT AltEA(S) lNVULVW1 &EWYEN
((I .
_________-_-___
_I__
HOSI’ITIL
.h.. Id. STlLLMAN
DAY *I, . , Y.U.
___-___-___-__--___________-_______-__-
Flgure 2. Sample surgical dtscharge summary as printed by computer.
There is at this time a set of 740 multiple choice questions (called the questionnaire) in the memory bank of the computer for the Trauma Service. Of these, a subset of 25 to 30 is used for any particular patient. This subset is determined by the nature of the injury, the areas involved, and the treatment and complications if any. If it is decided to enlarge the scope of inquiry in any area, additional questions may easily be added to the questionnaire immediately with no disruption of the rest of the program. A brief instruction sheet is available to the head of the department, enabling him, with no prior data processing experience, to program and modify the overall questionnaire [S]. This entire system is therefore self-sufficient, requiring no professional computer personnel. Any hospital can rent
such a computer remote terminal and immediately have MEDS at its disposal. Results
To date, 495 discharge summaries have been entered and reside in the memory of the computer. Only 5 to 10 minutes of the physician’s time and an amount of machine time costing an average of $1.41 were required for the entry of each one. This com-
lho American Jcumal ti Surgery
Medical Research by Computer
ABDOMINAL INJURIES AGE DISTWIEUTION
(XlLONIC
_______--
o-9 lo-
1 MYMYMYMMMMYWMHFFF FFF
-
29
;YYYYYYMYYYMMMMMYLlYMMRFFFFFFFFFRFFF :MYMYMHYYMFFFFFFFFFF
30
-
39
-
49
:YMYMYMMYYFFFF
xl
-
59
:YYWYF
60
-
69
:YYYF
10
-
79
80
-
89
90
-
99
IO
I5
20
COLONIC 25
30
35
40
45
____ __
__~~~~~_____~~~___~;~~~~ TAP
______---_------I
tr,
I-, LAYA(iz
T*P ,->
nITAL LAVAPE
Figure 3. Data of#a&ed retmqnWtively kom tha computer. An analpls of 95 abdominal injuries. Top, age and sex dlstrtftutfen. Bottom, comparison of operative ih?dlngs In patients grouped according to restdts of abdominal paracentesis and lavage.
pares favorably with the standard dictated, typed, and signed summaries, which cost our hospital $2.24 each. To determine the completeness of the discharge summary, we counted the number of discrete pieces of information (facts) in each of 100 dictated summaries selected at random from the files of our medical records department. While the average dictated summary contained 15 f 7 facts, the machine-generated summaries from these same charts were found to have 53 f 20 relevant facts. Despite this surfeit of information, the equivalent of only one line of typewritten text (about 81 characters) is required to store within the system the coded information for the average computerized discharge summary. MEDS develops effective summaries of the medical record to complement the original chart and serve as the basis for storage and retrieval of information. A duplicate printed discharge summary, written in concise and easily understandable English and comprehensible to professional and allied health personnel, can be obtained at any time in a matter of minutes. Performance of an automated clinical study is as simple as entering the question numbers you wish to
Volume 132, Aqusl1976
INJURIES
-
29 27 5 2 9 2
85.3 79.4 14.7 5.9 26.5 5.9
34
PATTERN-OF-CARE
REVIEW
COMPLICATIONS
50
PERCENT
“I>
HEVIEW
____- ____-_- ______-_ _____ ---___-_
SUTURE UEBWIkMENT !+ESECTION PR I HAkY ANASTI)MOS I S COLOSTUMY EXTEHIORIZATION ___-___-_-_-__-__-_--__________-_______ TOTAL PATIENTS
. . . . ..*....1.........*....l.....‘.........,....:....~ 5
PATTEHN-OF-CAGE - LT. COLON
______________________-_________L___
I9
20
40
INJURIES OPERAl’IOk
we__
-__-________---___-__^_______
N ______I
_-_______-q-w-
INTh-ABDOMINAL PULMONARY HENAL CAHD I AC GkNEHAL WOUND OTHER ______--___-__________________ TOTAL PAT1 mm
____
%
_----_---
6 44.4 7 38.9 3 16.7 I 5.6 6 33.3 II 61.1 I 5.6 --L-w 18
Flgvn, 4. San&& d&da obtabredlorPSROpaMm-ot-care review.(#atethatthesecasesweresekctedtromapatient population dftterent tiom that in Flgure 3.)
review. Figure 3 represents such a study performed on our Trauma Service. This “research project” ran less than 1 minute of computer time and included scanning of all 495 discharge summaries, selection of the 95 cases of abdominal trauma, and analysis of results. This information compares exploratory laparotomy findings with the results of abdominal paracentesis and lavage and agrees with previous studies in substantiating the utility of these diagnostic procedures [7]. The increased frequency of negative laparotomies in patients operated on with a positive tap or lavage is probably due to the fact that the diagnostic procedure itself was performed due to the difficulty of diagnosis based on clinical findings alone. Note that in no case did a patient with both negative tap and negative lavage require operation. That is, as expected, there were few falsepositive and no false-negative results. Although this study is derived from the discharge summaries stored in the machine’s memory at this time, with minimal effort it can be repeated in the future to provide upto-date statistics and allow the discovery and analysis of trends in our patient population. Comment
A manually reviewed clinical research project may take several months to perform. By the time the final data are obtained, the work often must be updated.
211
Stillman et al
The decision to embark on a retrospective medical study of significant magnitude therefore requires assurance that the results will be worthwhile. The reviewer is not personally familiar with each subject in the study and his biases may influence his interpretation of the often illegible and incomplete medical records. Because the essential information of all the charts is entered prospectioely on discharge into the computer without knowledge of the future application of the data, retrospective reviews are essentially unbiased. All statistics are calculated and analyzed by the machine, thus eliminating human prejudice. Our on-line data retrieval provides. a mechanism for periodic comparison of results and complications of different tests and procedures. Our computerized discharge summaries offer a unique opportunity to aid in the monitoring of PSRO guidelines. Individual cases can be easily retrieved for peer review using a number of screening criteria (length of hospital stay, investigations performed, number of complications or antibiotic usage). Pattern-of-care for any number of the common pathologic entities can be quickly analyzed without searching a myriad of paper. More importantly, due to the simplicity of reprogramming this system, the guidelines for screening may be periodically updated [8] and new clinical studies initiated. An example of the practical application of this retrieval ability is shown in Figure 4, which reveals a 33 per cent incidence of wound infection (11 of 34 cases) after left colon injuries. These data prompted us to initiate a prospective study in which the value of delayed primary skin closure after laparotomy for colonic injuries will be investigated. Summary
An adaptable on-line computer system for entry, retrieval, and analysis of medical discharge summaries has been developed and applied in the Trauma Service of a busy city hospital. Each summary occupied 5 to 10 minutes of the physician’s time and compared favorably in cost to the standard dictated summary. While the average dictated summary contained 15 f 7 relevant facts, the machine-generated summaries were found to have 53 f 20 relevant facts. The summaries are well organized, easily comprehensible, and a duplicate copy can be obtained at any time by using the patient’s name or number for identification. To date, 495 discharge summaries have been entered and reside in the memory of the computer. Statistical analysis from the data base is done by the computer, thus eliminating human prejudice. It provides facility for rapid
212
and accurate retrospective studies as well as pattern-of-care and individual case review. Acknowledgment: The authors would like to express their gratitude to John Stier, Computer Sciences Department, State University of New York at Stony Brook, and to Ronald L. Code, the Computing Center, Stanford University, for their encouragement and technical assistance. We thank Mrs Helen Berry and Mr William Goodman whose dedication have helped make implementation of the system possible. We are especially grateful to the interns and residents of the Trauma Surgery Service who together with the PDP-10 have created the actual data base from which the clinical information cited in this paper was derived. References 1. See VN: Screening in medical care assessment. Bull Am Co// Surg60: 7, 1975. 2. Buckley JD, Gledhill VX, Mathews JD, Mackay IR: SEARCH-A language for retrieval and statistical analysis of medical records. ComputBiomedRes 6: 235, 1973. 3. Davfs LS: A system approach to medical Information. M&h InformMed 12: 1, 1973. 4. Swedlow DB, Barnett Go, Grossman JH, Souder DE: A simple programming system (“Mver”) for the creation and execution of an automated medical history. Comf.xHBiomed Res 5: 90, 1972. 5. Stillman RM, Sawyer PN: Two medical applications of on-line data processing-research and education. Proceedings of 28th Annual ACEMB, 1975, p 28 1. 6. Stillman FM, Sawyer PN: MEDS-A simple interactive system for computerized medical discharge summaries. Med Instrumentation (To be published.) 7. Shaftan GW: Abdominal trauma, chapt 10. Practice of Surgery. Hagerstown, Maryland, Harper & Row, 1973. 8. Stolff JE: lntrahaspltal compliance with the PSRO law. NY State J h&d 75: 524, 1975.
Discussion Roy Cohn (Palo Alto, CA): As one whose sole experience with the computer is the angry patients who appear with a computerized bill that they have either paid or is in complete error, I have looked with a certain amount of skepticism upon the computer in research, believing that if research data are as inaccurate as the hospital billing program, the data bank may not be of very much value. It is obvious that the machinery is equal to the task, but that the programmer is the key, and as I listen to Doctor Kountz’s group, what they have described for their Trauma Service sounds excellent and is well worth trying in this particular problem. Many factors have created skepticism in the use of the computer for medical research. Probably the greatest one is the so-called “panacea syndrome” which, as Hofman has pointed out (Hospital Progress, April 1, 1971), has impeded more rapid progress. Many administrators have been burned by utopian systems that never delivered the predicted results. The most obvious cause for this syndrome is the zealous overselling by computer manufac-
The American Journal of Surgery
Medical Research by Computer
turers who have grossly underestimated the complexity of the research field, so that the competing groups who are exaggerating the computer’s ability have created even greater skepticism. An editorial in Hospitals in May 1967 quoted Ableson: “After growing wildly for years, the field of computing now appears to be approaching its infan-
quite straight-forward and it is reasonable to suspect that it will be successful so that it can serve as a model for future applications. It requires coordination among a minima1 number of personnel, demonstrates physical benefits, and can be accomplished within a reasonably short period of time.
cy."
In passing, it should be noted that the second major reason for the growth of the “panacea syndrome” is the status symbol consciousness of investigators. As an illegitimate status symbol, according to Hofman, the computer has proved expensive, not only in terms of the initial purchase or leasing expense, but also in terms of the costly impact it has had on personnel not adequately prepared for and involved in its implementation. In addition to alleviating the common fears associated with computers and investigating valid criticisms of specific applications, those responsible for instituting such a change should also anticipate concern for the depersonalization of patient care. One of the virtues of the proposed program is that it is
Samuel L. Kountz (closing): First, Doctor Cressman, I would like to thank you for awarding me the gold medal, the first gold medal I have ever won. Doctor Stillman and I will take this back to New York in the hope it will solve our fiscal problems. I appreciate Doctor Cohn’s remarks. Doctor Stilhnan has alleviated this problem of assuring accurate data input by providing the system with a certain degree of automatic error detection. In addition, its ease of operation makes it well accepted by our residents, and we feel it might solve some of our problems in quality control of care, and also in administration. I look at this as a means of determining, for example, how various residents performed as they rotated through the service.
213
With the expanding complexity of diagnostic and therapeutic modalities, the difficulty of performing an accurate retrospective study is increasing and hence vital decisions are often based on studies of relatively few patients while data on thousands of similar cases go unrecorded. Anyone who has ever attempted to complete a sizable chart review is altogether too familiar with the inevitable problems of handwritten, disorganized medical records. It is similarly frustrating to the physician to go over the chart of readmitted patients for it is often impossible to identify and interpret significant information. Incomplete, illegible or erroneous data, the unavoidable bias of the reviewer, and the prohibitive time required for an accurate study are responsible for distortion or loss of valuable clinical data. Each physician will have knowledge of his personal cases but will not have an overall appreciation of the results obtained by others. Ideally, the raw data for the retrospective study should be compiled prospectively by the physician responsible for the care of each patient. These problems in medical data processing now demand investigation and solution in order to allow implementation of the Professional Standard Review Organization (PSRO), which became a fact by Public Law 92-603. Assurance of high quality health care requires facilities for screening numerous parameters in each of many hospital admissions extending from concurrent review (while the patient is still hospitalized) through individual case review to retrospective pattern-of-care review [I].
From the Department of Surgery, State Univw~lly ofNew York, Downstate Medical Cater, and Kings County HostSSalCenter, Brocklyn, New York. This work was supported in part by the SNYMERF Fud. Reprint requests should be addressed to Richard M. Stillman. MD, Department of Surgery, Downstate Medical Center, Brooklyn, New York 11203. Presented at the Forty-Seventh Annual Meeting of the Pacific Coast Surgical Association, Monterey, California. February 15-18, 1976.
Volume 132, August 1976
For these reasons computerized medical charting has received increasing attention, and numerous systems have been designed to handle this formidable problem [Z-4]. Although used satisfactorily by their originators, most of the previous computer systems find little application outside their place of origin due to the complexity of programming for other specialities or difficulty of application in the hands of inexperienced users. In addition, the magnitude of the problem has required acceptance of often too strict limitations in the scope of information desired. We believe that the key to the solution of these problems lies in the discharge summary. While medical records as a whole document the detailed course of a patient’s hospitalization, discharge summaries should attempt to condense the material present on the chart without omitting major facts. Presented here is our experience withan adaptable computerized medical data system (MEDS) which was designed to assure not only simplicity of entry of the discharge summary and accuracy and speed in performing statistical analyses from the information stored, but also ease of actually programming the computer for eliciting the information required on any particular service [5J. Methods In a room adjacent to the Trauma Surgery Ward at Kings County Hospital there is a Hazeltine 2000 video screen terminal and printer (similar to that used by airlines at their ticket counters). This is connected by ordinary telephone lines to a large PDP-10 digital computer 40 miles away. This machine is available 24 hours a day for entry, retrieval, and analysis of discharge summaries. On the wall above the terminal is posted a short instruction sheet showing the user how to turn on the terminal, dial the computer, and initiate communication. On admission, the biographical data for each patient are entered into the computer by the secretary. On discharge,
209
Stillman et al CIHCUMSlANCE5 OF INJUicY I ASSAULT 2 MGTOh VEHICLE 3 ANIMAL ATTACK 4 SUICIDAL 5 FALL 0 MACHINEHY I SPOHTS 8 F.B.INGEST. 9 hOME ACCIOENT 10 OTHtH CH(X)SE ALL AI’PLICABLE 7 _I_
. .._ _. ..-....__... . .-
AGEhT I SHAHP INSTWUMENT 2 BULLET 3 SHUT GUN 4 BLUNT TIlAUKA 5 HUllAN BITE 6 ANIMAL BITE I MOTOR VEHICLE B F.B. 9 OTHER CHOOSE ALL APPLICABLE ? 2 BULLET (CALIBRE) = 7 & THAT SEEMS HATHEH HIGH. ARE YOU CEKTAIN BULLET IS 83 CALIBWE (YES OH NO) ? w BULLET (CALIBRE) = ? a
AREA(S) INVOLVED I HEAU 2 NECK 3 VERTtWHAE 4 CHEST 5 ABGOYEN 6 EXTREMITY I GYk U YELVIS I(T. 9 PELVIS LT. CH(X)SE ALL APPLICABLE
1HAT
_
/MISPITAL
CWWSE/
CONCOYlTrANT UIItASEI *ONE ANT‘B,OT,C “SAGE, PWWHYLACIIC, ,YH,GAr,ON OUTCOHEI COWLICATED CWPL.8 PULWNAHY (UN M)SP,TAL DAY I21 P,,LYONA,,Y U,YPLICATIO~S~ ATkLtCTh 15 “,SCkA”OE COND,T,ONI UISCHAWEU IHPRWE”
72
/ABDOMEN/
SIGNS
4,
Figure 1. Portion of a sampfe Interactfon during lfw 5 fo 10 minute Process of compllhtg a discharge summery for a surgical patlent. Users responses are under&ml for clarrtv.
entry by the physician
SLOW
lOStal *“LLEr 138 CALIRML) ENTnY WOUNDI “W EXlT 1)0UN08 SACK SYWTMS8 PAIN - LOCAL1 IZED LOCAL,,?E” O”A”“,NO. AHNOPMAL S”1EL SWWS SIGNS, LOCALILE” lENbE cNtSS. STUD,651 *SDw.Y)r*“. “I 1INALYbIS URINALYSISI NORMAL *BDOI,NAL WAY kES”L,l NORMAL AHEAS OF ASU.,NVOLVEDI ARD.NALL. DEF,NIrC ‘“TDA-aL)” ‘,A’ T,,EYACYl ,MMUIE”,ATE OYtWATION TlYt SElllEEY ADY,SS,Ok AhD SUltGE)tY* I-2 MS. TYPE ““E-W HEPLACEMENT: ll,!OLE RLWO (2 “NITS), CHYSTALLOI us oom cc, TYPE 1NT”A-W WEPLACtYENTl *HOLE BLOW L3 “NITS,. PLASMA , 250 cc,. CHfSTALLOIUS ti !ocm cc) .“* “F”IWNEAL P1LNET”ATIW. RLW” IN PEYITOhhAL CAVITY, FIkUlkGS AT L... ._ . . SO, I,, WOAN IN.,“PV OWCiirNm LIVEN LlVEk LO&Em ulGH1 LIVE” ,N.,U”Yl PUNCTUW’ ( SU LLET 1 A, TIME OF EYPLOWATTON LlVE” ,NJUHYI SLEEDlhG ~~X.L,YEHI smwt ~16.. “YA,N(S,USY”* SUMP AhD %%k’ NOM “EGHEt OF ARUOYINAL CM “TTAL,NAT,,,N*
_..___ ._.. -...
B ABlrOhMAL BONEL SOUNDS 9 OMENTAL HELENI AT ION IO BOkEL HEkNIATION II MASS 12 NOM OF TkSE CH(X)SE ALL APPLICABLE 7 2.
I
/AB”ONEN,
SYMPTOMS I PAIN - LOCAL I ZED 2 PAIN - GENEdALIZED 3 I’AIN - RiiFt~MEll TO SHOULDER 4 PAIN - RLFEtMED TO FACK :, HEMATEMES IS b VOMI’IING 7 NAUSEA 8 CONS1 IPATIOti 9 HtLENA IO NOnb OF TkSE CH(X)SE ALL APPLICARLE 7L
of the patient’s name or chart number provides the key to identification of the discharge summary and initiates a series of questions that appear on the screen. (Figure 1.) As the answer to each of these is typed, the machine checks it for validity and saves the response in a coded form. When the interaction is complete, a copy of the discharge summary is printed (Figure 2) for signature and insertion into the patient’s hospital chart. Confidentiality is guaranteed because a secret password is required to gain access to the stored information.
210
..~
TRAUMATIC@ EYEWENCY TlYE FROM INJUWY TO APWIVAL AT HOSPITAL (HOUWS,I UP TO CONDITION ON ADY,SS,OH* SHKX VITAL S,ONS YEWYUE” ON ADY,SS,ONr PULSE ,120 SEATS/YIN), PRESSUWE SYSTOLIC (00 MY/%). S.P.U,ASTOL,C (3” MWHG) RESUSCI’TATION‘ RLW3. CLEAV FLUID “ESYWSS TO WESUSCITATION~ GCOD CIUCUllSTANCES W lNJLl”Yr ASSAULT AltEA(S) lNVULVW1 &EWYEN
((I .
_________-_-___
_I__
HOSI’ITIL
.h.. Id. STlLLMAN
DAY *I, . , Y.U.
___-___-___-__--___________-_______-__-
Flgure 2. Sample surgical dtscharge summary as printed by computer.
There is at this time a set of 740 multiple choice questions (called the questionnaire) in the memory bank of the computer for the Trauma Service. Of these, a subset of 25 to 30 is used for any particular patient. This subset is determined by the nature of the injury, the areas involved, and the treatment and complications if any. If it is decided to enlarge the scope of inquiry in any area, additional questions may easily be added to the questionnaire immediately with no disruption of the rest of the program. A brief instruction sheet is available to the head of the department, enabling him, with no prior data processing experience, to program and modify the overall questionnaire [S]. This entire system is therefore self-sufficient, requiring no professional computer personnel. Any hospital can rent
such a computer remote terminal and immediately have MEDS at its disposal. Results
To date, 495 discharge summaries have been entered and reside in the memory of the computer. Only 5 to 10 minutes of the physician’s time and an amount of machine time costing an average of $1.41 were required for the entry of each one. This com-
lho American Jcumal ti Surgery
Medical Research by Computer
ABDOMINAL INJURIES AGE DISTWIEUTION
(XlLONIC
_______--
o-9 lo-
1 MYMYMYMMMMYWMHFFF FFF
-
29
;YYYYYYMYYYMMMMMYLlYMMRFFFFFFFFFRFFF :MYMYMHYYMFFFFFFFFFF
30
-
39
-
49
:YMYMYMMYYFFFF
xl
-
59
:YYWYF
60
-
69
:YYYF
10
-
79
80
-
89
90
-
99
IO
I5
20
COLONIC 25
30
35
40
45
____ __
__~~~~~_____~~~___~;~~~~ TAP
______---_------I
tr,
I-, LAYA(iz
T*P ,->
nITAL LAVAPE
Figure 3. Data of#a&ed retmqnWtively kom tha computer. An analpls of 95 abdominal injuries. Top, age and sex dlstrtftutfen. Bottom, comparison of operative ih?dlngs In patients grouped according to restdts of abdominal paracentesis and lavage.
pares favorably with the standard dictated, typed, and signed summaries, which cost our hospital $2.24 each. To determine the completeness of the discharge summary, we counted the number of discrete pieces of information (facts) in each of 100 dictated summaries selected at random from the files of our medical records department. While the average dictated summary contained 15 f 7 facts, the machine-generated summaries from these same charts were found to have 53 f 20 relevant facts. Despite this surfeit of information, the equivalent of only one line of typewritten text (about 81 characters) is required to store within the system the coded information for the average computerized discharge summary. MEDS develops effective summaries of the medical record to complement the original chart and serve as the basis for storage and retrieval of information. A duplicate printed discharge summary, written in concise and easily understandable English and comprehensible to professional and allied health personnel, can be obtained at any time in a matter of minutes. Performance of an automated clinical study is as simple as entering the question numbers you wish to
Volume 132, Aqusl1976
INJURIES
-
29 27 5 2 9 2
85.3 79.4 14.7 5.9 26.5 5.9
34
PATTERN-OF-CARE
REVIEW
COMPLICATIONS
50
PERCENT
“I>
HEVIEW
____- ____-_- ______-_ _____ ---___-_
SUTURE UEBWIkMENT !+ESECTION PR I HAkY ANASTI)MOS I S COLOSTUMY EXTEHIORIZATION ___-___-_-_-__-__-_--__________-_______ TOTAL PATIENTS
. . . . ..*....1.........*....l.....‘.........,....:....~ 5
PATTEHN-OF-CAGE - LT. COLON
______________________-_________L___
I9
20
40
INJURIES OPERAl’IOk
we__
-__-________---___-__^_______
N ______I
_-_______-q-w-
INTh-ABDOMINAL PULMONARY HENAL CAHD I AC GkNEHAL WOUND OTHER ______--___-__________________ TOTAL PAT1 mm
____
%
_----_---
6 44.4 7 38.9 3 16.7 I 5.6 6 33.3 II 61.1 I 5.6 --L-w 18
Flgvn, 4. San&& d&da obtabredlorPSROpaMm-ot-care review.(#atethatthesecasesweresekctedtromapatient population dftterent tiom that in Flgure 3.)
review. Figure 3 represents such a study performed on our Trauma Service. This “research project” ran less than 1 minute of computer time and included scanning of all 495 discharge summaries, selection of the 95 cases of abdominal trauma, and analysis of results. This information compares exploratory laparotomy findings with the results of abdominal paracentesis and lavage and agrees with previous studies in substantiating the utility of these diagnostic procedures [7]. The increased frequency of negative laparotomies in patients operated on with a positive tap or lavage is probably due to the fact that the diagnostic procedure itself was performed due to the difficulty of diagnosis based on clinical findings alone. Note that in no case did a patient with both negative tap and negative lavage require operation. That is, as expected, there were few falsepositive and no false-negative results. Although this study is derived from the discharge summaries stored in the machine’s memory at this time, with minimal effort it can be repeated in the future to provide upto-date statistics and allow the discovery and analysis of trends in our patient population. Comment
A manually reviewed clinical research project may take several months to perform. By the time the final data are obtained, the work often must be updated.
211
Stillman et al
The decision to embark on a retrospective medical study of significant magnitude therefore requires assurance that the results will be worthwhile. The reviewer is not personally familiar with each subject in the study and his biases may influence his interpretation of the often illegible and incomplete medical records. Because the essential information of all the charts is entered prospectioely on discharge into the computer without knowledge of the future application of the data, retrospective reviews are essentially unbiased. All statistics are calculated and analyzed by the machine, thus eliminating human prejudice. Our on-line data retrieval provides. a mechanism for periodic comparison of results and complications of different tests and procedures. Our computerized discharge summaries offer a unique opportunity to aid in the monitoring of PSRO guidelines. Individual cases can be easily retrieved for peer review using a number of screening criteria (length of hospital stay, investigations performed, number of complications or antibiotic usage). Pattern-of-care for any number of the common pathologic entities can be quickly analyzed without searching a myriad of paper. More importantly, due to the simplicity of reprogramming this system, the guidelines for screening may be periodically updated [8] and new clinical studies initiated. An example of the practical application of this retrieval ability is shown in Figure 4, which reveals a 33 per cent incidence of wound infection (11 of 34 cases) after left colon injuries. These data prompted us to initiate a prospective study in which the value of delayed primary skin closure after laparotomy for colonic injuries will be investigated. Summary
An adaptable on-line computer system for entry, retrieval, and analysis of medical discharge summaries has been developed and applied in the Trauma Service of a busy city hospital. Each summary occupied 5 to 10 minutes of the physician’s time and compared favorably in cost to the standard dictated summary. While the average dictated summary contained 15 f 7 relevant facts, the machine-generated summaries were found to have 53 f 20 relevant facts. The summaries are well organized, easily comprehensible, and a duplicate copy can be obtained at any time by using the patient’s name or number for identification. To date, 495 discharge summaries have been entered and reside in the memory of the computer. Statistical analysis from the data base is done by the computer, thus eliminating human prejudice. It provides facility for rapid
212
and accurate retrospective studies as well as pattern-of-care and individual case review. Acknowledgment: The authors would like to express their gratitude to John Stier, Computer Sciences Department, State University of New York at Stony Brook, and to Ronald L. Code, the Computing Center, Stanford University, for their encouragement and technical assistance. We thank Mrs Helen Berry and Mr William Goodman whose dedication have helped make implementation of the system possible. We are especially grateful to the interns and residents of the Trauma Surgery Service who together with the PDP-10 have created the actual data base from which the clinical information cited in this paper was derived. References 1. See VN: Screening in medical care assessment. Bull Am Co// Surg60: 7, 1975. 2. Buckley JD, Gledhill VX, Mathews JD, Mackay IR: SEARCH-A language for retrieval and statistical analysis of medical records. ComputBiomedRes 6: 235, 1973. 3. Davfs LS: A system approach to medical Information. M&h InformMed 12: 1, 1973. 4. Swedlow DB, Barnett Go, Grossman JH, Souder DE: A simple programming system (“Mver”) for the creation and execution of an automated medical history. Comf.xHBiomed Res 5: 90, 1972. 5. Stillman RM, Sawyer PN: Two medical applications of on-line data processing-research and education. Proceedings of 28th Annual ACEMB, 1975, p 28 1. 6. Stillman FM, Sawyer PN: MEDS-A simple interactive system for computerized medical discharge summaries. Med Instrumentation (To be published.) 7. Shaftan GW: Abdominal trauma, chapt 10. Practice of Surgery. Hagerstown, Maryland, Harper & Row, 1973. 8. Stolff JE: lntrahaspltal compliance with the PSRO law. NY State J h&d 75: 524, 1975.
Discussion Roy Cohn (Palo Alto, CA): As one whose sole experience with the computer is the angry patients who appear with a computerized bill that they have either paid or is in complete error, I have looked with a certain amount of skepticism upon the computer in research, believing that if research data are as inaccurate as the hospital billing program, the data bank may not be of very much value. It is obvious that the machinery is equal to the task, but that the programmer is the key, and as I listen to Doctor Kountz’s group, what they have described for their Trauma Service sounds excellent and is well worth trying in this particular problem. Many factors have created skepticism in the use of the computer for medical research. Probably the greatest one is the so-called “panacea syndrome” which, as Hofman has pointed out (Hospital Progress, April 1, 1971), has impeded more rapid progress. Many administrators have been burned by utopian systems that never delivered the predicted results. The most obvious cause for this syndrome is the zealous overselling by computer manufac-
The American Journal of Surgery
Medical Research by Computer
turers who have grossly underestimated the complexity of the research field, so that the competing groups who are exaggerating the computer’s ability have created even greater skepticism. An editorial in Hospitals in May 1967 quoted Ableson: “After growing wildly for years, the field of computing now appears to be approaching its infan-
quite straight-forward and it is reasonable to suspect that it will be successful so that it can serve as a model for future applications. It requires coordination among a minima1 number of personnel, demonstrates physical benefits, and can be accomplished within a reasonably short period of time.
cy."
In passing, it should be noted that the second major reason for the growth of the “panacea syndrome” is the status symbol consciousness of investigators. As an illegitimate status symbol, according to Hofman, the computer has proved expensive, not only in terms of the initial purchase or leasing expense, but also in terms of the costly impact it has had on personnel not adequately prepared for and involved in its implementation. In addition to alleviating the common fears associated with computers and investigating valid criticisms of specific applications, those responsible for instituting such a change should also anticipate concern for the depersonalization of patient care. One of the virtues of the proposed program is that it is
Samuel L. Kountz (closing): First, Doctor Cressman, I would like to thank you for awarding me the gold medal, the first gold medal I have ever won. Doctor Stillman and I will take this back to New York in the hope it will solve our fiscal problems. I appreciate Doctor Cohn’s remarks. Doctor Stilhnan has alleviated this problem of assuring accurate data input by providing the system with a certain degree of automatic error detection. In addition, its ease of operation makes it well accepted by our residents, and we feel it might solve some of our problems in quality control of care, and also in administration. I look at this as a means of determining, for example, how various residents performed as they rotated through the service.
213
