Portfolio item: Coursework paper (S513, Organizational Informatics, Fall 2012) & business brief

"Normal accidents in Medicine: The system is to blame"

Abstract & Introduction

The normal accident is a frequent occurrence in fields ranging from aviation to medicine. In such accidents, the individuals involved do follow standard operating procedures (albeit with seemingly minor mistakes); these minor mistakes nonetheless result in major morbidity (suffering) and mortality (death) among victims. The humans involved in these accidents receive much of the blame for their minor errors and resulting accidents, but those who place the blame tend to ignore profound errors in the systems that share responsibility in mediating these accidents. Such misattribution is particularly common with respect to normal accidents in the medical field. In this analysis, after reviewing literature on general normal accidents, lay news articles on medically-related normal accidents, and medical literature on relevant conditions and medical technologies, I discuss two cases of normal accidents in medicine: One of a young infant given a fatal overdose of cardiac medication, and one of a boy who died due to undiagnosed septic shock. I then discuss some potential reasons these accidents occurred, with an emphasis on systematic errors that pervade the medical system rather than individual human error. Furthermore, I propose potential solutions to avoid these types of normal accidents; these solutions range from minor changes in existing communication structures to the implementation of entirely novel technologies, and are largely based upon suggestions and conclusions made from the literature. Finally, I will speculate on future ramifications these cases (and their potential solutions) bring to the medical field.

Literature Review

Perrow (1999) gives an excellent introduction to the idea of a normal accident. In particular, Perrow's concept of the "system accident" (p. 70) describes many medically oriented normal accidents rather well. In such a system accident, extremely complex systems, combined with natural human limitations, may cause a normal system accident (p. 72). Weick (1990) discusses a normal accident not in the medical field, but one in aviation. Weick's analysis of the 1977 Tenerife Air disaster nonetheless cites issues that are common to both the medical and aviation fields, including tightly coupled systems and human fatigue (pp. 121-125).

Human fatigue may be a factor in overdose error. Defined as the injurious administration of too high of a dose of medication to a patient, it is a medical form of normal accident and a common source of morbidity and mortality in hospital patients. In the case to be studied here, the overdose involved digoxin, a cardiac medication whose toxic dose is relatively close to its therapeutic dose (Belkin, 1997). Therefore, a slight error in the prescribed dosage can easily be fatal to the patient; slight errors such as this decimal point error are particularly likely when physician handwriting is of poor quality. One manufacturer of digoxin, GlaxoSmithKline (GSK), states that the highest acceptable single adult dose of digoxin is 0.6mg (milligrams); in the case in question, this amount was exceeded by 50% (GlaxoSmithKline, 2009). Related to dosing issues are e-scribing systems, a form of electronic medical records (EMRs). These systems take input from the physician as an explicit medication and dosage; identical information is sent to the dispensing pharmacy. Such systems receive typed prescriber input, which is then transmitted to the pharmacy. Depending on the e-scribing system used, dosages that are not standard require special authorization from the physician in the form of an additional on screen dialog confirmation (Tamblyn et al, 2012).

Ofri (2012) cites anchoring bias as another major cause of medical error; anchoring bias may be seen as one of many pervasive cultural issues pertaining to the authority of the physician. Anchoring biases have the capability to cause dosing errors if physicians assume that what is written on a prescription (no matter how erroneous) is correct; these types of errors are notorious for causing overt misdiagnosis. In general, diagnoses are treated as discrete entities; missed diagnoses in atypical disease presentations are commonplace, creating a phenomenon known as anchoring bias (Ofri, 2012).

Sepsis, also known as septic shock, is a runaway type reaction the human body may have in response to overwhelming infection, and may be difficult to diagnose due to anchoring bias (Dwyer, 2012). Its symptoms, similar to that of ordinary infections, therefore anchor the physician to the diagnosis of a milder infection and not sepsis (Dwyer, 2012). Left untreated, it quickly causes irreversible organ damage and then death. The overwhelming infections leading to sepsis are frequently caused by major penetrating trauma, such as deep or widespread cuts in the skin. However, the Stop Sepsis Collaborative has identified these common signs as diagnostic of sepsis (United Hospital Fund, 2012). EMR systems that are similar to the aforementioned e-scribing systems have the potential to automatically alert clinicians to signs of potentially fatal conditions given if such signs are provided (Hafner, 2012). However, it must be noted that EMR systems may require long periods of time to fully justify themselves in medical organizations; this slowness causes wariness among all parties involved in EMR adoption (Lohr, 2009). Nonetheless, for reasons of governmental regulations, finances, and safety, many medical organizations are rapidly embracing EMRs. In particular, Gawande (2007) cites intensive care situations as appropriate and beneficial places in which to institute a medical system known as the checklist; this type of system is similar to the United Hospital Fund's (2012) proposition, and certainly amenable to implementation in EMR systems.

Tightly coupled systems, where one action is highly dependent upon the previous action are seen in these cases; these systems create a chain of command that is vulnerable to a domino effect of unanticipated consequences (Gusterson, 2011; Perrow, 1999; Weick, 1990). Belkin (1997) cites the fact that all clinicians, regardless of experience level, do in fact make mistakes; these mistakes are often due to poor understanding of one's own cognitive abilities (p. 44). Similarly, Weick (1990) also implicates tightly couple systems and human fatigue as causes of normal accidents. A lack of EMR usage and automated EMR error checking is also cited as a cause of major medical errors. When EMRs are not checked, errors of egregiously incorrect dosage (Tamblyn, 2012) as well as missed diagnoses are more probable (Hafner, 2012).

Furthermore, the tightly coupled structure of the medical system is also known for its lack of horizontal linkages (Daft, p. 106) in terms of communication. Similarly, in hospital systems, nurses and residents hold a lower position in the organization and generally obey physician orders, regardless of any personal objection that might be present. Evidence of such fear has been shown to exist between residents and attending physicians (Friedman et al, 2010). In addition, these systems may be ignorant of both themselves and their own environments (Goebel et al, 2012). As such, doctors may ignore nurses, patients, and patients' families; Goebel et al. particularly emphasize the lack of communication between general practitioners and hospitals. The horizontally linked structure is also difficult to implement in the cultural sense, as deep and pervasive cultural changes in the organization are required (Daft, p. 111). Therefore, we can safely say that this non-horizontal culture is vital in creating what appear to be normal accidents.

Specific case studies: The stories of Jose Martinez (1997) & Rory Staunton (2012)

One of these normal accidents occurred in 1997: Jose Martinez, a young Houston infant with a cardiac defect, was prescribed the cardiac medication digoxin in order to ensure his survival until surgery on the defect could be performed (Belkin, 1997). However, the intern responsible for writing the prescription prescribed ten times the dose of digoxin that he had intended to prescribe to the small infant Martinez; the intended dose was 0.09 milligrams (mg), while the dose was written out as 0.9mg. The attending physician did not notice the resident's error, but both the dispensing pharmacist and an attending nurse did. (Belkin, p. 4). While the pharmacist did attempt paging the resident to notify him of the error, the resident was by this time off duty, and the nurse resorted to questioning another resident about the errant dosage. The second resident himself misread the incorrect dosage as correct, presumably as a 0.9mg was visually similar to 0.09mg. The nurse, although "troubled" (p. 4) by the ordered dosage, administered the drug to Martinez, who then rapidly died.

A similar systems based error occurred in 2012 with Rory Staunton, a twelve-year-old boy in New York. After suffering a cut on his arm during a basketball game, Staunton fell ill with symptoms of a generalized infection as bacteria present on his skin invaded the cut, giving him a blood infection. His pediatrician did not attribute the infection related symptoms to the cut Staunton received on his arm and informed the family that the boy was suffering from an unrelated malady. Later on, Staunton developed more serious medical symptoms, and was referred by his pediatrician to the New York University (NYU) hospital for further investigation. Much like the pediatrician, the clinicians at NYU did not attribute his condition to the cut in his arm (which would have been the only reason to consider blood infection and consequent septic shock as the cause of his problems), and therefore failed to recognize the ailment as such. Soon after his discharge from NYU, Staunton developed a high fever and was again seen by his pediatrician, who again referred him to NYU. By this point, signs of clear septic shock had developed, with Staunton's bodily organs failing from his body's response to the overwhelming bacterial infection. The organ failure proved irreversible, and Staunton died soon after his re-admission to NYU (Dwyer, 2012).

Analysis

Much as some in the aviation field view the Tenerife air disaster as virtually unavoidable (Weick, 1990), we may be tempted to think that both of these deaths were acceptable casualties of the current medical system. I argue in contrast that the deaths were due to failures in the structure and function of the medical system and therefore potentially avoidable in current times. While dosing errors are not uncommon, severe ones (such as the tenfold overdose that proved fatal to Martinez) are relatively rare. In the case of Martinez, one must note that the difference between the 0.09mg dose Martinez required and the 0.9mg dose prescribed is ten-fold; these dosages are visually similar but strikingly different in terms of effects, the former being therapeutic and the latter being fatal. The tightly coupled chain of command, resident within a vertical communication structure, had few places in which individuals would be convinced to check for and act upon these types of errors: In Martinez's case, the pharmacist, attending physician, attending nurse, and two medical residents had access to the improperly written dosage. While the pharmacist and nurse both caught the error, the prescription appeared normal to the attending physician and the second resident, both of who held higher positions on the chain of command and were therefore able to allow the fatal overdose to be administered. Lastly, of note is that there was no electronic system present at the time to apprehend the prescribing error.

Therefore, I conclude that Martinez's case was a systematic error in the chain of command and in communication, which may have been aggravated by a biased routinization seen in the medical field and the lack of adequate electronic surveillance. In effect, an anchoring bias was created (Ofri, 2012), as the nurse assumed whatever erroneous dosage written on the prescription was correct, as it originated from a physician. In Staunton's case, the behaviors of both a major medical hospital and Staunton's own pediatrician were factors in his death, and also acted upon anchoring bias. Staunton's case was not one of a misplaced decimal point; nonetheless, it was still a case of systematic medical errors that were strikingly similar to those seen in Martinez's case.

Furthermore, like Martinez's case, Staunton's case also showed a strong lack of horizontal communication and inadequate usage of EMRs along with poor communication between his pediatrician and the NYU hospital. It is also seen that the clinicians did not take into account Staunton's parents' suspicions regarding his condition (Belkin, 1997). Much as in the case of Martinez, the communication structure present in Staunton's case magnified the physicians' anchoring bias. The linear and discrete process of diagnosis constituted an anchoring bias (Ofri, 2012) that was tightly coupled (Perrow, 1999), as Staunton's physicians essentially ignored alternate diagnoses largely due to the fact that these diagnoses were not obvious to them initially (Dwyer, 2012). It must be noted that in Staunton's case, EMRs were also not used to their full capability as outlined by Hafner (2012). Much as Gawande (2007) suggests, utilization of a checklist may have saved Staunton's life, and such a checklist could be integrated into an EMR system.

Recommendations & Outlook

Methods of preventing of these normal accidents are therefore obvious. EMR systems known as e-scribing systems have the potential to reduce fatal prescribing and dosage errors (Tamblyn, 2012). GlaxoSmithKline (2009) states that there is a clearer way to write the dosage: The standard recommended in more modern times is to write digoxin dosage in micrograms (ug) and not milligrams, where one milligram is one thousand micrograms. We may thus conclude that such integration of e-scribing with proper drug information has the capability to facilitate the recognition and prevention of dosing errors. Similarly, more conventional EMR systems should contain features to check for potential missed diagnoses, as physicians, like all human beings, are not perfect in terms of their capability during the course of their jobs. In Staunton's case, the system could have noted the boy's history of a previous laceration, combined this note with the aberrant vital signs that were observed, and alerted the physicians to a potential problem (Hafner, 2012).

There also exists impetus for social change: The control structures that underpin medical systems must be reviewed in context, and horizontal linkage and communication capabilities must be added. As such, the potential exists to loosen the tightly coupled system present and make communications within it more horizontal than vertical. In a more loosely coupled system, the "troubled" (Belkin, p. 4) nurse in Martinez's case would have had equal weight in the decision to administer the dosage and need not have feared sanctions for disobeying physician orders and refusing to administer the fatal dose. In Staunton's case, the rights and opinions of the patients and his or her family ought to have been taken into consideration with equal weight as to physician observations; it is clear that Staunton's parents had insight on their son's medical conditions. In future cases that may arise, if horizontal communication is better implemented, physicians may in fact take into consideration opinions of the patient and the patient's family. Horizontal linkage in the form of better communication between Staunton's pediatrician and the NYU hospital would have also been of benefit to the situation.

In terms of current beliefs, one may view the situation as bleak: Due to the high amount of bureaucracy and vertical structure present in medical organizations, implementation of these systems in a thorough fashion will likely be slow. Horizontal structures are difficult to implement when the existing organizational structure is more vertically oriented (Daft, p. 111). Some authors also consider the rate of EMR adoption as poor (Lohr, 2009), and when medical organizations approach EMR adoption, they may do so slowly and incompletely. Nonetheless, due to the long-term benefits of EMR adoption and horizontal structuring, it is hoped that over the coming several years, we will see fewer normal accidents such as the ones experienced by Martinez and Staunton. Adoption of new technologies, electronic and otherwise, has already shown benefit in the medical field (Gawande, 2007). Furthermore, implementation of advanced-warning systems and horizontal communication has the possibility to reduce the rate of normal accidents in all fields.

References

Belkin, L. (1997, 15 June). How can we save the next victim? New York Times. Retrieved from ProQuest Historical Newspapers, 1857(SM28-).

Daft, R. (2001). Organizational theory and design. 7th Edition. Cincinnati, OH: Southwestern Publishing.

Dwyer, J. (2012, 11 July). An infection, unnoticed, turns unstoppable. New York Times. Retrieved from http://www.nytimes.com/2012/07/12/nyregion/in-rory-stauntons-fight-for-his-life-signs-that-went-unheeded.html

Friedman, S., Sowerby, R., Guo, R., Bandiera, G. (2010). Perceptions of emergency medicine residents and fellows regarding competence, adverse events, and reporting to supervisors: a national survey. Canadian Journal of Emergency Medicine, 12(6), 491-499. Retrieved 2012, November 28 from http://www.cjem-online.ca/v12/n6/p491

GlaxoSmithKline Inc. (2009). Prescribing information: Lanoxin [digoxin] (Monograph). Retrieved 2012, November 30 from Food and Drug Administration: www.accessdata.fda.gov/drugsatfda_docs/label/2010/009330s025lbl.pdf

Gawande, A. (2007, December 10). The checklist. The New Yorker. Retrieved from http://www.newyorker.com/reporting/2007/12/10/071210fa_fact_gawande

Goebel, B. et al. (2012). Stakeholder perspectives on handovers between hospital staff and general practitioners: an evaluation through the microsystems lens. British Medical Journal: Quality and Safety, 21, 106-113. Retrieved 2012, November 30 from http://qualitysafety.bmj.com/content/21/Suppl_1/i106.long

Gusterson, H. (2011, 16 March). The lessons of Fukushima. Bulletin of the Atomic Scientists. Retrieved 2012, December 1 from http://www.thebulletin.org/web-edition/columnists/hugh-gusterson/the-lessons-of-fukushima

Hafner, K. (2012, December 3). For second opinion, consult a computer? The New York Times. Retrieved from http://www.nytimes.com/2012/12/04/health/quest-to-eliminate-diagnostic-lapses.html

Lohr, S. (2009, November 16). Little benefit seen, so far, in electronic patient records. New York Times. Retrieved from http://www.nytimes.com/2009/11/16/business/16records.html

The New York Times. (2012). Medical records in sepsis death. Retrieved from http://www.nytimes.com/interactive/2012/07/11/nyregion/medical-documents-in-sepsis-death.html

Ofri, D. (2012, July 19). Falling into the diagnostic trap. New York Times. Retrieved from http://well.blogs.nytimes.com/2012/07/19/falling-into-the-diagnostic-trap/

Perrow, C. (1999). Normal Accidents. 2nd Edition. Princeton, NJ: Princeton University Press.

Tamblyn, R., Reidel, K., Patel, V. (2012). Physicians' responses to computerized alerts for psychotropic drugs in older persons. British Medical Journal Open, 2012(2). Retrieved from http://bmjopen.bmj.com/content/2/5/e001384.long

Weick, K. (1990). The vulnerable system: An analysis of the Tenerife air disaster. Journal of Management, 16(3), 571-593 [supplied as pp. 117-144].

United Hospital Fund. (2012). STOP sepsis collaborative. Retrieved from http://www.uhfnyc.org/initiatives/quality_improvement/STOP_Sepsis